Remote Sensing Applications: Society and Environment 35 (2024) 101218

Contents lists available at ScienceDirect

Remote Sensing Applications: Society and

Environment
journal homepage: www.elsevier.com/locate/rsase

A review on hyperspectral imagery application for lithological

mapping and mineral prospecting: Machine learning techniques
and future prospects
Soufiane Hajaj a, *, Abderrazak El Harti a, Amin Beiranvand Pour b, Amine Jellouli a,
Zakaria Adiri c, Mazlan Hashim d
a Geomatic, Georessources and Environment Laboratory, Faculty of Sciences and Techniques, Sultan Moulay Slimane University, Beni Mellal, Morocco
b Institute of Oceanography and Environment (INOS), Higher Institution Center of Excellence (HICoE) in Marine Science, University Malaysia Terengganu
(UMT), Kuala Nerus 21030, Malaysia
c Laboratory of Geo-heritage, Geo-Environment and Mining and Water Prospecting, Faculty of Sciences, Mohammed Premier University, Oujda, Morocco
d Faculty of Built Environment & Surveying, Universiti Teknologi Malaysia UTM, Johor Bahru 81310, Malaysia

ARTICLE INFO ABSTRACT

Keywords: Hyperspectral remotely sensed imagery is a pertinent instrument in lithological and mineral al-
Hyperspectral remote sensing imagery terations mapping for a vast range of ore mineralization. This imagery typically provides an ideal
Alteration minerals characterization and exploitation of the Earth's outcrop, using wide-ranging spectral/spatial data
Lithological mapping for the reconnaissance stages during ore mineral exploration. The application of hyperspectral re-
Machine learning algorithms
mote sensing datasets derived from satellite and airborne platforms has proven to be instrumental
Mineral prospectivity mapping
in surmounting prevalent challenges encountered in mineral exploration endeavors. Because of
the exponential surge in hyperspectral remote sensing data acquisition from disparate platforms,
the scientific community has been incited to develop sophisticated and resilient data processing
approaches using artificial intelligence (AI) techniques. Additionally, recent studies have wit-
nessed the integration of machine learning (ML) algorithms with conventional image processing
techniques and geological surveys, featuring the upward significance of hyperspectral remote
sensing lithological mapping and mineral prospecting (HLM-MP). Although there are previous re-
views that broached the use of HLM-MP, there is still a lack of an updated comprehensive review
on the subject. This review article unequivocally demonstrates the potential inherent in harness-
ing hyperspectral imaging datasets and ML algorithms, facilitating precise mapping of crucial ge-
ological features and enabling the production of significantly enhanced mineral prospectivity
mapping. Furthermore, this review identifies promising prospects for the utilization of deep
learning (DL), multisource data integration, and cloud computing when processing hyperspectral
remote sensing data, thereby further refining HLM-MP investigations.

1. Introduction
Hyperspectral data acquisition involves capturing a series of continuous and contiguous spectral bands using a remote sensing sen-
sor, that is cautiously designed to represent diagnostic absorption and reflection characteristics of the Earth's surface materials
(Bedini, 2012; Clark and Rencz, 1999). Hyperspectral imagery (HSI) demonstrates great capability as a valuable tool in a wide range

* Corresponding author.
E-mail address: [email protected] (S. Hajaj).

https://doi.org/10.1016/j.rsase.2024.101218
Received 2 December 2023; Received in revised form 19 March 2024; Accepted 3 May 2024
Available online 6 May 2024
2352-9385/© 2024 Elsevier B.V. All rights reserved.
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

of Earth observation (EO) applications, such as geological mapping, geography, agriculture, climate change monitoring, forestry, mil-
itary, security sectors, etc. (Rast and Painter, 2019; Shimoni et al., 2019; Transon et al., 2018). Hyperspectral remote sensing provides
detailed spectral information across a wide range of wavelengths (i.e., the visible and near-infrared (VNIR) and the short-wave in-
frared (SWIR)), enabling precise identification and characterization of mineralogy and lithological compositions (Cloutis, 1996).
However, operative processing and analysis of hyperspectral data pose significant challenges due to its complexity and high dimen-
sionality. Nowadays, AI techniques, including ML and DL algorithms, have demonstrated great potential in handling and extracting
valuable information from hyperspectral datasets (Pour et al., 2021; Shirmard et al., 2022). Innovative ML algorithms and image pro-
cessing techniques allow the extraction of mineralogical and lithological information in complex metallogenic provinces and inacces-
sible zones using HSI (Hajaj et al., 2023b; Pour et al., 2021; Shirmard et al., 2022). The capability of ML and DL algorithms, as well as
the level of their computational power, enable exploration geologists to successfully overcome the challenges faced at various phases
of mineral exploration campaigns (Asadzadeh and de Souza Filho, 2016; De La Rosa et al., 2021; Xu et al., 2022). Consequently, the
incorporation of HSI and AI techniques is a cutting-edge and worthwhile approach for future prospects of lithological mapping and
mineral exploration communities.
The VNIR - SWIR (0.4–2.5 μm) and the thermal infrared (TIR) (8–12 μm) wavelength ranges provide information to remotely de-
tect the mineralogical composition of outcropping lithological units (Hunt, 1977, 2017). That is to say, the VNIR spectral bands are
able to detect and characterize the iron oxides/hydroxides; for instance, VNIR bands of HyMap hyperspectral sensor were successfully
used to identify the spectral characteristics of Ca-bearing silicates containing ferrous (Fe2+) and ferric (Fe3+) ions in Ca-Skarn hosted
U-REE mineralization (dos Reis Salles et al., 2017). Additionally, Al–OH, Fe–OH, and Mg–OH features and vibrations combination are
mainly responsible for spectral characteristics of alteration minerals in the SWIR region (Hunt, 1977, 2017). Al–OH (e.g., phylosili-
cate minerals) and Mg–Fe–OH/CO32− (e.g., carbonate, chlorite, and epidote) alteration minerals associated with ore mineralization
can be detected using hyperspectral imaging sensors’ SWIR bands (Bedini, 2017). Fig. 1 shows HyMap, EnMAP, PRISMA, AVIRIS-NG,
AHSI, AISA, and Hyperion sensors bands positions in 0.4–2.5 μm with typical spectral features for igneous, sedimentary and meta-
morphic rock units. Additionally, the TIR region is useful for detecting rock-forming minerals such as quartz, amphibole, and
feldspar. This is mainly caused by the existence of basic Si–O bond vibrations. For instance, Thermal Airborne Spectrographic Im-
agery (TASI) TIR data were successfully mapped lithological units, quartz, microcline, diopside, calcite, and hornblende (Liu et al.,
2021).
Despite the fact that there are commercial software packages available for calibrating, processing, and validating the results of
mineral mapping from hyperspectral data, various levels of customization still needed to be improved using innovative techniques.
These variations are typically due to the occurrence of unpredictability in Earth's surface and atmospheric conditions during data ac-
quisition. Nonetheless, signal-to-noise ratio estimation (Meer et al., 2001), data dimensionality reduction (Green et al., 1988), end-
member extraction (Meer et al., 2001), and mapping/classification process (Manolakis et al., 2001) are generally considered as the
main steps of preprocessing/processing hyperspectral data in mineral mapping.
Previous review articles have discussed geological hyperspectral remote sensing from various perspectives. Cloutis (1996) evalu-
ated the analytical techniques used in geological hyperspectral remote sensing, where, he highlighted the necessity of intelligent HS
data processing techniques for extracting valuable information using reduced data volume. Van der Meer et al. (2012) reviewed the
geological hyperspectral (with multispectral) remote sensing studies; then, they have principally raised challenges related to valida-
tion techniques, moreover, they concluded that most of those studies do not reveal interdisciplinary, and multidisciplinary character-
istics. Bedini (2017) reviewed the application of geological hyperspectral remote sensing to various typologies of ore deposits study

Fig. 1. Continuous HyMap, EnMAP, PRISMA, AVIRIS-NG, AHSI, AISA, and Hyperion sensors bands positions in 400–2500 nm with typical spectral features (solid area)
for each of the igneous, sedimentary, and metamorphic rocks (Hunt, 2017).

2
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

and exploration. Peyghambari and Zhang (2021) attempted to perform an updated review of the geological hyperspectral imaging
techniques, expecting to develop advanced techniques in the future; however, they stated that the available HSI data for analysis was
limited.
The lithological mapping and mineral prospecting via HSI remote sensing has witnessed a significant surge in research activities,
with an extensive body of literature dedicated to HSI in mapping lithology (more than 145 papers) and mineralogy (over 1000 pa-
pers) over the past six years. This increase in studies underscores the growing importance and interest in leveraging hyperspectral
data for geological applications. Notably, the recent deployment of advanced hyperspectral satellite sensors has addressed the histori-
cal challenge of limited global availability, marking a transformative shift in the accessibility of high-resolution spectral data for re-
searchers worldwide. To provide a comprehensive overview of the current state of hyperspectral geology research, the current up-
dated review will not only synthesize the studies conducted over the past six years but also undertake a critical examination of the re-
cent literature. This examination will allow to highlight the concurrences and divergences with existing reviews, shedding light on
the evolving perspectives within the field. Accordingly, points of agreement, unresolved questions, methodological advancements,
and research trends as well as future prospects have been discussed. In the first place, a set of hyperspectral platforms have been re-
viewed, in the second place the recent studies in HLM-MP have been reviewed with the corresponding used techniques for endmem-
ber extraction, dimensionally reduction (DR), enhancement, lithological classification, and mineral prospectivity mapping, in the
third place, some research trends for optimizing HLM-MP have been discussed including band selection (BS), HSI Unmixing, Cloud
computing, and data integration. Lastly, discussions and conclusions have been extracted. The current review highlights the frequent
use of spectral algorithms compared with a lack of exploiting ML and DL algorithms for unlocking the full potential of HSI in LM-MP
studies.

2. Literature synopsis on hyperspectral geological mapping over the last six years
Scientific contributions were investigated through Scopus, Science Direct and Web of Science for the years 2018–2023 (last six
years), using some keywords such as “hyperspectral” and “lithology” or “lithological” and also “hyperspectral” and “mineral” or
“mineralogical”. Indeed, the distinction between mineral and lithological mapping has been performed by searching each of them
separately and plotting the results in 3D-diagram as shown in Fig. 2. Fig. 2A and B illustrate the tendency of contributions in lithologi-
cal and mineral mapping using hyperspectral sensors for the years 2018–2023. A huge difference between the number of manuscripts
published in hyperspectral mineral mapping and hyperspectral lithological mapping was observed. As evident from the presented
plots, there has been a steady rise in the volume of scientific publications dedicated to the utilization of hyperspectral technology in
lithological and mineral mapping over the last six years. For example, in hyperspectral lithological mapping, during the period
1986–2017, the number of published documents is 1229, while in the last six years only, 1090 documents have been published (Fig.
2A and B). Similar progress can be remarked with hyperspectral mineral mapping, with the evident difference between the amounts
of published documents by year. During the period 1996–2017, the number of published documents was 128, while in the last six
years only, 142 documents have been published (Fig. 2B). Fig. 2C shows the number of publications using the keywords of satellite,
airborne, and UAV with “hyperspectral” and “lithology” or “lithological” on one side, as well as with “hyperspectral” and “mineral” or
“mineralogical” on the other side. The difference between the numbers of publications in hyperspectral lithological over mineral
mapping is obvious in Fig. 2C, too. Fig. 2C demonstrates that the most used platforms in hyperspectral lithological and mineralogical
mapping are satellite and airborne. Furthermore, the contributions using satellite and airborne sensors reveal a comparable number
over the last decade for each category (accordingly in lithology or mineralogy). Fig. 2D shows the number of publications using the
keywords “Hyperspectral” in combination with “Lithology” or “Mineralogy” or “Mineral” or “Exploration” as a function of their high-
est ten funding sources. The National Natural Science Foundation of China (NNSFC) National Key Research, the Development Pro-
gram of China (NKRDPC), and the National Aeronautics and Space Administration (NASA) are ranked as the first three funding
sources of researches using hyperspectral remote sensing techniques for mineral prospection and lithological mapping over the last
six years.

3. Hyperspectral remote sensing platforms

The choice of platform relies upon the nature of the specific application. The platforms may include satellites, airborne, unmanned
aerial systems, or ground-based systems. Table 1 shows the characteristics of the main hyperspectral spaceborne, airborne, and
ground-based imager sensors. Additionally, Fig. 3 shows hyperspectral sensors succession at different altitude levels. The latter ex-
hibits the importance of unmanned Aerial Systems (UAS) in fulfilling the altitude gap observed between ground and airborne plat-
forms.

3.1. Spaceborne platform

Several spaceborne hyperspectral sensors have been instrumental in advancing our understanding of Earth's surface geology. One
notable example is Hyperion, which was launched aboard NASA's Earth Observing-1 (EO-1) satellite in 2000. Hyperion provides 242
spectral bands in the range of 0.356–2.577 μm. Therefore, several studies for geological exploration aims were performed using Hy-
perion data. Nevertheless, these data present some issues, including limited spatial coverage, the existence of strips, and the smile ef-
fect in some Hyperion scenes (Sun et al., 2008). Recently, PRISMA was developed by ASI (Agenzia Spaziale Italiana) with 250 bands
(in 0.4–2.5 μm spectral range). This sensor provides a spatial resolution of 30 m and a spectral resolution of 12 nm. Also, the German
Spaceborne Imaging Spectrometer Mission EnMAP has been launched in space in 2022. EnMAP presents high radiometric, spectral
accuracy, and stability. The current investigations using EnMAP focus more on environmental monitoring (Chabrillat et al., 2023),

3
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 2. (A and B) The last six years' documents published in the Hyperspectral lithological and mineral mapping field (Scopus). (C) Number of published docu-
ments categorized with platform and application. (D) Main funding organizations of the mineral/lithological mapping studies during the last six years. Abbrevi-
ations: Horizon 2020 Framework Programme = H2020FP; Fundamental Research Funds for the Central Universities = FRFCU; Natural Sciences and Engineer-
ing Research Council of Canada = NSERCC; National Science Foundation = NSF; Chinese Academy of Sciences = CAcademyAS; National Key Research and
Development Program of China = NKRDPC; National Aeronautics and Space Administration = NASA; National Natural Science Foundation of
China = NNSFC, China Postdoctoral Science Foundation=CPSF, Ministry of Science and Technology of the People's Republic of China=MSTPRC.

whereas, this sensor with a number of 224 bands in VNIR and SWIR regions can provide a promising result in geological applications
(Chabrillat et al., 2021). With PRISMA and EnMAP sensors, geological and mineralogical mapping would be more reliable and ex-
haustive. In turn, AHSI sensor onboard the GaoFen-5 satellite has 330 spectral bands in the 0.4–2.5 μm spectral range. Some investi-
gations were conducted in environmental monitoring and mineral exploration using AHSI data (e.g., Dong et al., 2022; Sun et al.,
2023). These sensors capture detailed spectral information across a wide range of wavelengths, enabling precise identification and
characterization of materials on Earth's surface. The data collected by these sensors have revolutionized several research fields as
mineral prospecting, land cover mapping, ecosystem monitoring, and other scientific endeavors, providing invaluable insights into
our planet's complex dynamics.

4
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Table 1
Characteristics of the main hyperspectral spaceborne, airborne, and UAV-based sensors. Abbreviations: National Aeronautics and Space Administration = NASA;
European Space Agency = ESA; Digital Airborne Imaging Spectrometer = DAIS; Hyperspectral Mapper = HyMap; Compact High-Resolution Imaging Spectrome-
ter = CHRIS; Hyperspectral Imaging SUIte = HISUI; Advanced Visible-Infrared Imaging Spectrometer Next Generation = AVIRIS-NG; Japanese Ministry of Econ-
omy, Trade, and Industry = METI; PRecursore IperSpettrale Della Missione Applicativa = PRISMA; Environmental Mapping and Analysis Program = EnMAP; Hy-
perspectral InfraRed Imager = HyspIRI; Gaofen-5 = GF-5; Hyperspectral Digital Imagery Collection Experiment = HYDICE; Airborne Imaging Spectrometer for Ap-
plications=AISA; German Aerospace Center=DLR; China Aerospace Science and Technology Corporation=CASC; National Center of Space Studies = CNES; Multi-
User System for Earth Sensing = MUSES; Earth Sensing Imaging Spectrometer Hyperspectral=DESIS.

Sensor type Sensor Number of bands Platform Spatial Spectral Launch Main applications
(spectral range μm) (Organization) resolution resolution date

Spaceborne Hyperion 242 (0.356–2.577) EO-1 (NASA) 30 m 10 nm 2000 Earth observation, lithological and mineral
mapping.
CHRIS 63 - 150 (0.4–1.05) PROBA-1 (ESA) 25–50 m 1.25–11 nm 2001 Water, soil, aerosol studies.
PRISMA 250 (0.4–2.5) PRISMA (ASI) 30 m 12 nm 2019 Atmosphere, Natural resources
management.
EnMAP 228 (0.42–2.45) EnMAP (DLR) 30 m 5 and 10 nm 2022 Environment monitoring, earth's resources
management.
DESIS 235 (0.4–1)) MUSES (DLR) 30 M 2.55 nm 2018
HISUI 185 (0.4–5.5) ALOS-3 (METI) 30 m 10–12.5 nm 2020 Vegetation, energy management.
AHSI 330 (0.4–2.5) GF-5 (CASC) 30 m 5–10 nm 2018 Vegetation, Earth observation, Mineral
mapping.
HypXIM 192 (0.4–0.25) HypXIM (CNES) 4–8 m 10 nm 2018 Geosciences, vegetation, coastal
ecosystems, security and defense
EMIT 285 (0.38–2.5) ISS (NASA) 60 m 7.4 nm 2022 Earth's mineral dust.
Airborne HyMap 126 (0.45–2.48) (Integrated 1–10 m 10–20 nm 1998 Mineral mapping, and, Vegetation.
Spectronic)
AVIRIS 224 (0.38–2.51) (NASA) 20 m 10 nm 1986 Mineral mapping
AVIRIS-NG 425 (0.38–2.51) (NASA) 5m 10 nm 2015 Geology, vegetation, water, and
atmosphere.
HyspIRI >200 (0.38–2.50) HyspIRI 60 m 10 nm 2020 Natural disasters.
CASI-1500 288 (0.38–1.05) (ITRES) 1m 2.3 nm 1988 Forestry, land-use planning, agriculture,
aquatic monitoring, and mineral mapping.
SASI-600 100 (0.95–2.45) (ITRES) 2–2.5m 15 nm Forestry, land-use planning, agriculture,
aquatic monitoring, and mineral mapping.
HYDICE 210 (0.4–2.5) (Naval research 1m 10.2 nm 1994 Mineral mapping, bathymetry, water
Lab) quality, and vegetation.
ProSpecTIR 360 (0.39–2.45) (SpecTIR LLC) 1m 5 nm Mineral mapping, vegetation, climate
change, land use, and, other anthropogenic
impacts.
TASI 32 (8–11.5) (ITRES) 2.25 m 0.1091 nm 2009 Geology.
AISA 178 (0.4–2.24) (ITRES) 2m 10 nm 1993 Geology.
DAIS 79 (0.4–14) (GER 20 m 0.9–25 nm 1995 Mineral mapping.
Corporation)
AisaFENIX Up to 348 (0.38– (SPECIM) – 0.5–14 nm – Environmental Monitoring, Mining
1K 2.5) (Geological Exploration) and Agriculture
AisaOWL 100 (7.7–12) (SPECIM) 2m 45 nm – Geological exploration, and Target
detection.
UAV-based Mjolnir VS- 490 (0.4–2.5) (HySpex) – 3–5.1 nm – Vegetation, and others.
620
MicroCASI 200 (1–2.5) (ITRES) – – – Vegetation and related studies.
Nano- 270 (0.4–1) (Headwall) – 6 nm – Environmental Monitoring, and Mining.
Hyperspec
Aisa (0.4–1) (SPECIM) – 1.75/3.5/7 nm – Mapping and monitoring vegetation.
KESTREL 10

Fig. 3. Relative importance of UAS hyperspectral remote sensing over Earth surface detection.

5
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

3.2. Air-borne imaging spectroscopy

Hyperspectral airborne sensors have proven to be invaluable tools in geology, enabling detailed characterization and mapping of
geological features and materials. These sensors capture spectral information across a wide range of wavelengths, allowing geologists
to identify and analyze the composition and mineralogy of the Earth's surface with remarkable spatial precision (which is advanta-
geous compared to the majority satellite-based sensors). One prominent example is the Airborne Visible/Infrared Imaging Spectrome-
ter-New Generation (AVIRIS-NG), a widely used hyperspectral sensor in geology (e.g., Guha et al., 2021b; Kumar et al., 2020a;
Tripathi and Govil, 2019a). With a spatial resolution of 8.1 m, AVIRIS-NG provides 427 contiguous bands, spaced at 5 nm intervals
spanning the wavelength range of 0.38–2.51 μm. AVIRIS-NG was developed as an enhanced version of its predecessor, AVIRIS, incor-
porating several modifications and improvements in its design, offering superior data quality, characterized by a high signal-to-noise
ratio (SNR), and absence of keystone and smile/spikes errors due to the use of a triple-blaze convex grating (Chapman et al., 2019;
Tripathi et al., 2020). Among these sensors-type, there is also HyMap, which is a high-resolution hyperspectral sensor (5 m), provid-
ing 126 spectral bands in the range 0.45–2.48 μm. It was launched in 1998 by Integrated Spectronics, Sydney Australia, and operated
by HyVista Corporation. Since its launch, HyMap has demonstrated a great capability in mineralogical detection and mapping studies
(Bedini, 2012; Bedini et al., 2009; Hajaj et al., 2023a, 2023b; Molan et al., 2014). Another notable sensor is the HYDICE sensor pre-
senting 210 bands in 0.4–2.5 μm. Ground surface sampling ranging between 0.75 and 3.75 m was attributed to HYDICE data. Cuprite
Mining District, Nevada is one of the most known mineralogically investigated areas using the latter's data (Resmini et al., 1997). One
notable example is the AisaFENIX 1K characterized by 420 spectral bands, ranging between 0.375 and 2.5 μm, with a 1.5 m spatial
resolution. AisaFENIX can be used ground-based or with an airborne platform. AisaFENIX 1K has been used in various geological ap-
plications, including mapping lithological units and mineral alterations (Meyer et al., 2021). These hyperspectral airborne sensors en-
able geologists to detect subtle variations in spectral signatures associated with different geological formations, aiding in the under-
standing of geologic processes. By providing detailed spectral information at a fine spatial resolution, hyperspectral airborne sensors
revolutionize geologic investigations and contribute to advancing our knowledge of Earth's geology. Some of the most important air-
borne hyperspectral sensors from various organizations are represented in Table 1.

3.3. UAS and ground-based imaging spectroscopy cameras

Unmanned Aerial Systems (UAS) equipped with hyperspectral sensors have revolutionized geological research by providing high-
resolution and flexible data acquisition capabilities. In the context of geologic lithology and mineralogy application, the UASs play a
pivotal role in high-resolution EO applications by providing hyperspectral data characterized by both exceptional spatial and tempo-
ral resolution (Lodhi et al., 2018). Fig. 3 illustrates the conceptual bridging of altitude gap that exists between traditional airborne
platforms and ground-based methods, highlighting the unique capabilities offered by UAS in this context. An example is the HySpex
Mjolnir VS-620 sensor, which offers hyperspectral imaging capabilities in both the VNIR and SWIR regions. Mjolnir VS-620 mounted
on UAS platforms have been employed in geological investigations for mineral exploration and environmental monitoring (Thiele et
al., 2021) (see Table 1). Some of the most important hyperspectral UAS cameras from various organizations are presented in Table 1.
These instruments offer high spatial and spectral resolutions, enabling geologists to accurately identify and map minerals, rocks, and
alteration zones (Pour et al., 2023; Shahmoradi et al., 2020).
Ground-based imaging spectrometers have emerged as crucial tools in geology for acquiring detailed spectral information about
geological features and materials. Data collection is carried out at ground level utilizing a hyperspectral imaging system, which is af-
fixed to either a moving or static platform. The static platform, in particular, exhibits superior observation capabilities and temporal
resolution, rendering it an advantageous choice for data acquisition in various scenarios (Lodhi et al., 2018). In generally, such sen-
sors are suitable in mining due to their located analysis capability (Gladwell et al., 1983). Many sensors (e.g., AisaFENIX) can be used
both ground-based or installed on specific airplanes, a set of UAS-based sensor can be used ground-based (Lodhi et al., 2018). In addi-
tion to ground-based sensors, within the domain of hyperspectral imaging (HSI) laboratory analysis, the SisuROCK drill-core scanner
workstation manufactured by Specim, Spectral Imaging Ltd, and equipped with hyperspectral sensors (FX10e, SWIR, FENIX, and
FX50), has been widely employed for drill-core HSI applications (De La Rosa et al., 2021). For instance, SisuROCK equipped with
Specim SWIR used along with an LWIR-HS hyperspectral camera have been successfully used to identify hydrothermal alteration fea-
tures in drill cores from the Canadian Malartic gold deposit (Lypaczewski et al., 2019). Krupnik and Khan (2019) undertook a review
focused on close-range, ground-based hyperspectral imaging within the context of mining applications. To sum up, in other to provide
a global view about the HSI sensors, the characteristics of some of the main hyperspectral sensors are synthetized and presented in
Table 1, further details about HSI sensors have been reviewed extensively by the authors (Adão et al., 2017; Agar and Coulter, 2007;
Lodhi et al., 2018; Lu et al., 2020).

4. Hyperspectral imagery in geological mapping

4.1. Spectral dimensionality reduction and enhancement techniques
The hyperspectral data have redundant spectral information due to their contiguous spectral channel, which will require a long
computation time to be processed. To optimize processing time while preserving valuable information, data can undergo dimension-
ality reduction. Dimensionality reduction (DR) methods are classified into distinct categories, including unsupervised, and supervised
techniques, as well as linear, and non-linear approaches. Among the unsupervised DR approaches, the ICA, PCA, and MNF stand out
as the most widely employed methods (Tripathi and Garg, 2021). In order to enhance lithological units based on hyperspectral data,
DR techniques were widely used. Indeed some of the main methods (i.e., PCA, ICA, Linear Discriminant Analysis (LDA) (Kale et al.,

6
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

2017), have been used either in the extraction of the target informative bands for specific lithological units, or as a pre-step in litho-
logical classification using AI algorithms (ML and DL classifiers). Table 2 shows the current studies that have used DR methods to hy-
perspectral data for lithological and mineral mapping aims. In the Otanmäki ore deposit, in Finland, Lorenz et al. (2021) studied the
performance of SVM classification based on different DR approaches for drone-borne VNIR data. The novel techniques, namely or-
thogonal total variation component analysis (OTVCA) and wavelet-based sparse reduced-rank regression (WSRR), as well, conven-
tional approaches such as PCA, and MNF were assessed. Consequently, the classification results show relative outperformance of the
novel methods implemented in this study. Adiri et al. (2020) showed that the implementation of the ICA to the data fusion of multi-
spectral (Sentinel-2A and Landsat-8 OLI) and Hyperion datasets could provide optimum mineralogical mapping results for detecting
chlorite, calcite, and malachite compared to Mixture Tuned Matched Filtering (MTMF) method. The spectral Pan-sharpening method
of color normalization (CN) was adopted in the multispectral-hyperspectral data fusion. However, ICA separates information even if it
only takes up a small portion of the image, which was attributed to it principle of data independence. Additionally, DLR’DESIS, and
PRISMA hyperspectral satellite with global coverage have been processed by DR techniques for lithological mapping in the study per-
formed by Tripathi and Garg (2021). The lithological/mineral mapping results show the suitability of the used sensors for geological
applications, as well as their possible application for large-scale mapping.
The RDB enhancement method is a simple (band math) method based on the three-point formulation ratio, where the addition of
the bands displaying the absorption shoulders was divided by the band representing the minimum of the absorption feature (Crowley
et al., 1989). Whereas, the Crosta method allows to create a new set of uncorrelated bands named PC bands, based on a selection of
bands from the original image (Crosta and Moore, 1990). Rani et al. (2020) used the RBD technique for distinguishing calcareous
(dolomite), argillaceous (phyllite), and siliceous (quartzite) (Table 2). By comparing the index image composite with the reference
lithological map provided by the Geological Survey of India and field validation data, the effectiveness of AVIRIS-NG data in accu-
rately mapping metasedimentary units within the Aravalli Supergroup was demonstrated (Guha et al., 2021b). The RBD was utilized
to effectively identify surface mineralization (Table 2), establishing a robust connection between rock types and their corresponding
surface proxies of mineralization in Bhilwara district, India. Mineral exploration implications were deduced from the study results,
where the correlation between rock types and their corresponding surface indicators of mineralization was established. The findings
of the study were supported by X-ray diffraction (XRD) analysis, geophysical, and geochemical data. Guha et al. (2022) developed
three RBDs (Table 2) derived from the image spectra of different rocks collected from AVIRIS-NG data. The false color composite
(FCC) image of these indices allowed to attain a comprehensive identification of many rocks in the Sittampundi layered complex, In-
dia. Furthermore, additional lithological mapping results were derived from the Crosta and Support Vector Machine (SVM) methods.
From the same bands used in the RBD method, three PC images were used in FCC to identify the study area rock types. Hajaj et al.
(2023a) developed three RBDs to map Al–OH, Mg–Fe–OH/CO3, and hydrothermal alteration minerals within Ameln Valley Shear
Zone (AVSZ), Morocco (Table 2), then applied Crosta technique to the used HyMap scene. Thereafter, the HyMap-derived layers (i.e.
Relative Absorption Band Depth (RBD), Crosta technique, MTMF, and Line density) were fused to generate a high-resolution mineral
potential map.
The PRISMA satellite data was used effectively to build several spectral indices, Laukamp (2022) used several spectral indices
based on PRISMA to map the bedrock (mafic), and the regolith (chlorite-epidote, Fe oxy/hydr-oxides, kaolin). The results comparison
to airborne HSI-derived geological mapping in the area, demonstrating the efficiency of PRISMA to detect regolith/bedrock. Cardoso-
Fernandes et al. (2022) applied enhanced methods based on band math to PRISMA in order to detect associated topaz to Li deposits
(Cornwall, UK). Chirico et al. (2022), PRISMA VNIR-SWIR satellite data, complemented to ground sample analyses, facilitated the ac-
curate mapping of dolomitization in the Jabali area; PRISMA enabled the clear differentiation between dolomites and limestones, and

Table 2
Current hyperspectral studies focused on using dimensionality reduction and enhancement methods in lithological and mineralogical mapping.

Method Dataset Feature RBD Formula RBD Algorithm (μm) References

RBD AVIRIS- Dolomite (B373+B404)/B392 (2.2397 + 2.3949)/2.3348 Rani et al. (2020)

NG Phyllite (B340+B366)/B348 (2.0744 + 2.2046)/2.1144
Quartzite (B353+B378)/B366 (2.1395 + 2.2647)/2.2046
RBD PRISMA Mg–OH (B149 + B154)/B152 (2.327 + 2.364)/2.349 Chirico et al. (2022)
Al-sheetsilicate, (B125 + B136)/ B133 (2.143 + 2.229)/2.206
Dolomites (B145 + B151)/B148 (2.298 + 2.342)/2.320
Fe oxides/hydoxides (B44 + B62)/B57 (0.770 + 9.62)/0.908
RBD AVIRIS- Calcite Not specified (2.184 + 2.389)/2.239 Guha et al. (2021b)
NG Sericite/illite Not specified (2.144 + 2.284)/2.204
Goethite Not specified (0.767 + 1.222)/0.937
RBD/SPCA HyMap Al–OH (B101+B111)/B108 (2.094 + 2.268)/2.217 Hajaj et al. (2023a)
Mg–Fe–OH/CO3 (B111+B120)/B114 (B111+B120)/B114
Fe3+/Fe2+ ((B21+B51)/B31)* ((0.755 + 1.205)/0.9)*
(B21/B4) (0.755/0.496)
RBD/SPCA AVIRIS- Anorthosite (B360+B372)/B366 (2.17 + 2.23)/2.20 Guha et al. (2022)
NG TTG (B361+B333)/B344 (2.179 + 2.03)/2.09
Metagabbro/Pyroxenite (B121+B111)/B113 (0.977 + 0.927)/0.942
MNF AVIRIS- Various rock types in the ultramafic – – Tamilarasan et al.
NG terrain (2022)
ICA Hyperion Calcite, malachite, and chlorite – – Adiri et al. (2020)

7
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

also facilitated the identification of gossan outcrops overlaying the mineralized areas rich in Mg–OH, Al-sheet silicate, dolomites, and
Fe oxides/hydroxides. Indeed, the developed spectral indices (Table 2) are principally based on the targeted minerals criteria, and lit-
erature study (Gaffey, 1986; Laukamp, 2022; Laukamp et al., 2011).

4.2. Endmember extraction technique in HSI

In geological hyperspectral remote sensing, the main approaches for selecting geological targeted materials endmembers for fur-
ther classification process are either classical, from outcrop or laboratory, or image pixels. In mineral exploration, researchers typi-
cally undertake either laboratory-based or field-based spectral parameter characterization about the targeted material, which lever-
ages prior geological records to establish background knowledge about the materials to be mapped. However, when Endmembers and
the scene are acquired under distinct conditions of solar irradiance, it becomes necessary to perform continuum removal to normalize
spectra on the average background reflectance, thus mitigating discrepancies arising from varying illumination conditions (Guha,
2020). For HSI pixel-derived endmembers, the conventional approach for endmember identification involves distinguishing image
pixels based on their spectral attributes. Van Der Meer (2000) emphasizes the critical significance of meticulous endmember selection
when employing unmixing techniques to avert singularity and orthogonality issues that may arise during the matrix inversion of lin-
ear systems. Veganzones and Grana (2008) categorized endmember extraction methods into Geometrically Oriented (e.g., F-
FINDER), Lattice Computing, and Heuristic (e.g., Pixel Purity Index (PPI)) methods. Some of the common endmember extraction
techniques are the PPI, Automatic Target Generation Process, (ATGP) N-FINDR, Vertex Component Analysis (VCA), and Iterative Er-
ror Analysis (IEA) (Zortea and Plaza, 2009). The PPI was integrated into the standardized methods developed by analytical imaging
and geophysics (AIG). Fig. 4 displays the schema of the hyperspectral analysis standardized procedure AIG. PPI is a heuristic end-
member extraction method widely used in mineral mapping studies (Agrawal et al., 2022, 2023). In the study conducted by Samani et
al. (2021), PPI was applied to the AVIRIS-NG derived minimum noise fraction (MNF) bands covering the Ambaji–Deri Region (India).
Hajaj et al. (2023a) used PPI for 13 HyMap-derived MNF bands to map hydrothermal alteration minerals in the Ameln valley shear
zone, Anti-Atlas, Morocco. Then, several minerals and mineral mixtures were detected and fatherly mapped using a sub-pixel algo-
rithm. Jain and Sharma (2019) applied PPI to AVIIRS-NG data for mineralogical mapping in Southeastern Rajasthan, India. There-
after, the sub-pixel classification results allowed a good discrimination of mineralized zones from settlements. George et al. (2022)
conducted a study comparing various spectral reduction techniques before applying endmember extraction to AVIRIS-NG data, cover-
ing the Indian sub-continent. Their study aims to evaluate the effectiveness of these techniques combined with endmember extraction
for LSU. The Principal Component Analysis (PCA), Independent Component Analysis (ICA), MNF, and Partial Informational Correla-
tion (PIC) feature selection approaches (e.g., Dabiri and Lang, 2018; Fassnacht et al., 2014; Paul and Kumar, 2019; Xia et al., 2013)

Fig. 4. Diagram of the hyperspectral analysis standardized procedure AIG (Boardman and Kruse, 1994).

8
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

were analyzed. The PIC method coupled with the N-FINDR endmember algorithm, demonstrated better results in pixel reconstruction
accuracy and computational efficiency.

4.3. Lithological discrimination and mapping

4.3.1. Lithological units’ classification
For lithological classification, per-pixel algorithms are more suitable and more adopted by authors than sub-pixel algorithms. The
lithological units relatively larger size than mineral phases can explain the mentioned choice suitability (Guha, 2020). Accordingly,
the studies using sub-pixel algorithms are relatively limited. For instance, Feng et al. (2018) employed LSU for lithological mapping
utilizing AISA (SWIR) and SEBASS (LWIR) airborne data (the application of spectral unmixing to HIS is more detailed The Section
5.2). Notably, significant discrepancies were observed for quartz-rich sediments, with the SWIR tending to overestimate their distrib-
ution, while the LWIR produced more consistent results in comparison to the existing maps. The presence of pervasive lichen coatings
on mafic rocks (basalt and gabbro) affected both SWIR and LWIR imagery, with SWIR yielding better outcomes than LWIR. To address
limitations encountered in independent data sets, the authors combined spectral data, resulting in accurate mapping of all geologi-
cally meaningful units, aligning with existing geological maps. Table 3 presents the current studies that implemented various per-
pixel and sub-pixel spectral-based, ML and DL algorithms for lithological classification.
Guha et al. (2022) used AVIRIS-NG imagery to apply SVM classification on a specific set of bands derived from PCA, the PCA data
set demonstrated accurate classification results than whole image data set, as well as the data set derived from the developed spectral
indices (Table 3). The method using limited hyperspectral spectral bands can be used to generate the comprehensive lithological
mapping for celestial bodies, such as the moon and other planets. Kumar et al. (2020b) demonstrated a higther performance of the
SVM in comparison to various other ML models. By employing the JMIM-derived optimal bands, the SVM model achieved remark-
able results with an Overall Accuracy (OA) of 85.48% and a Kappa Coefficient (KC) of 0.83. The utilization of JMIM-based optimal
bands proved to be particularly effective in accurately classifying a range of lithological units, including metabasalt, amphibolite,
granite, acidic intrusive, and migmatite, outperforming alternative input datasets. Fang et al. (2022) proposed a combinitional algo-
rithm of PCA, quantum-behaved particle swarm optimization (QPSO), and least-squares support vector machines (LSSVM) (PCA-
QPSO-LSSVM). This novel model was trained based on 32 PCA-derived bands from TASI imagery. The QPSO was applied to optimize
the kernel function σ2 and regularization parameter γ. Classification results of PCA-QPSO-LSSVM for TASI scene covering LiuYuan
town (GanSu province, China) demonstrate more accuracy in comparison with the simple LSSVM algorithm. Libeesh and
Arivazhagan (2023) used per-pixel and sub-pixel classifiers including Spectral Angle Mapper (SAM), LSU, and MTMF on PRISMA im-
agery encompassing the economically significant Nilgiri block in India. Sub-pixel algorithms have exhibited a higher level of preci-
sion in accurately delineating lithological variations within the study region, showcasing numerous mineral occurrences of Au, Pd,
Pt, and Cr. These algorithms have proven especially effective in classifying mafic and ultramafic rock types, namely pyroxenite, peri-
dotite, and gabbro. The validation of classification outcomes was accomplished through laboratory-based spectra, XRD analysis, and
comprehensive fieldwork, collectively reaffirming the substantial potential of hyperspectral imaging (HSI) in conjunction with sub-
pixel classifiers.
Thannoun (2021) used the SAM classification technique for lithological/mineralogical mapping. The mapping process involved
the initial extraction of endmembers from a Hyperion image using the PPI and N-Dimension visualization method. The identified end-
members were then utilized for training the SAM classification. The outcomes of this study highlight the significant capabilities of Hy-
perion data in effectively depicting the spatial distribution of alteration minerals and lithological units. These outcomes have mineral
exploration implications within a specific section of the Kirkuk anticline (Northern Iraq). Additionally, Cai et al. (2020) employed the
SAM to Gaofen-5 (GF-5) hyperspectral data to perform lithological mapping for a geologically complex mountainous region. Subse-
quently, a specific assessment approach was utilized to evaluate the accuracy of the mapping outcomes. This study furnished valuable
thematic insights for regional lithological investigations and offers substantial technical support for decision-making and surveys.
Lorenz et al. (2021) investigated SVM classification with novel feature extraction (FE) algorithms, where, studied the signifi-
cance of FE algorithms as a fundamental step to mitigating noise and reducing data complexity aiming to lithological classification.
This study examined the suitability of conventional (such as dimensionality reduction techniques) and innovative FE methods (Or-
thogonal Total Variation Component Analysis - OTVCA, and Wavelet-based Sparse Reduced-Rank Regression - WSRRR) through
three distinct geological HS mapping scenarios, encompassing thermal HS lithological mapping, terrestrial paleoseismic outcrop
analysis, and drone-based mineral prospection. The study underlines the novel feature extraction algorithms’ strength in both the
accuracy of classification and their ability to provide more geologically interpretable results.
Several DL categories were applied for HSI classification including, Convolutional Neural Network (CNN) (Yu et al., 2017), Auto
Encoder (AE) (Zhu et al., 2017), Recurrent Neural Network (RNN) (Mou et al., 2017), Deep Belief Network (DBN) (Ayhan and Kwan,
2017), and Generative Adversarial Network (GAN) (Zhu et al., 2018). The application of DL for hyperspectral lithological mapping is
typically focused on its ability in extracting the deep features data, which is valuable in handling surface geology variety and com-
plexity. Zhao et al. (2020) proposed a hierarchical spatial-spectral feature extraction with long short-term memory (HSS-LSTM) to
fuse the spatial-spectral feature for hyperspectral lithological classification. These features are initially derived by combining spectral
data with local spatial features from a CNN model, then LSTM efficiently captures correlations and acquiring hierarchical features.
The analysis of experimental findings from Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) hyperspectral data in the well-
known Nevada mining area, and the comparison between HSS-LSTM, SAM, 3D-CNN, and, LSTM, revealed that HSS-LSTM demon-
strated superior performance. Liu et al. (2021) studied three distinct CNN algorithms, comprising 1D-CNN, 2D-CNN, and 3D-CNN,
which were employed and evaluated in conjunction with the six relevant state-of-the-art ML methods. Lithology classification utiliz-
ing CNN was evaluated on thermal infrared hyperspectral data acquired from the TASI sensor at three distinct sites in Liuyuan, Gansu

9
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Table 3
The current hyperspectral studies in lithological mapping. Abs: Overall accuracy: OA.

Method Dataset Feature Classes number OA Research Main Highlights Reference

(%)

SVM AVIRIS-NG Igneous rocks: Meta-gabbro, 3 87.07 The SVM classification based on limited Guha et al.
anorthosite, and TTG. PC-derived bands provided accurate (2022)
classification results in mapping
Sittampundi layered complex, India.
1D-CNN TASI Three localized sites within 3 sets of 5 >90 The using of 3D-CNN with hyperspectral Liu et al. (2021)
Liuyuan, situated in Gansu thermal infrared data demonstrated
Province, China. promising results in small-scale
lithological classification.
1D-CNN HyMap Different parts around the Ameln 3 sets with 5, 7 >95 Processing airborne high-resolution Hajaj et al.
valley, western Anti-Atlas. and 7 hyperspectral data by 1D-CNN allowed (2023b)
Including, Proterozoic basement to accurate mapping of lithological units in
infra-Cambrian lithofacies. a semi-arid region.
XGB, SVM, RF PRISMA Ophiolitic rocks, metasedimentary 7 >90 Suitability of RF, XGB, and SVM was Shebl et al.
and metavolcanic rocks, granites, demonstrated in providing acceptable (2023)
and sedimentary rocks lithological mapping to update
lithological mapping data in Egyptian
desert region.
XGB ZY1-02D Marble, monzonitic granite, syenite 8 83.43 A new two-layer extreme gradient Lin et al. (2023)
granite, diorite, gabbro, granite boosting (XGBoost) model based on the
porphyry, and granodiorite, XGBoost decision tree and an enhanced
greedy search method was built for
lithological classification.
LSU AISA and Mafic, ultramafic, metamorphic, 9 – The outcomes of unmixing for both Feng et al. (2018)
SEBASS and sedimentary rocks from individual SWIR and LWIR data, as well
Nunavik, Quebec. as the combined data, demonstrate
distinct advantages when employing the
CWA (continuous wavelet analysis)
approach pre-processing step for HSI.
SVM AVIRIS-NG Classify lithological units in Hutti, 7 85.48 The SVM application for a JMIM-derived Kumar et al.
(JMIM India (including metabasalt, dataset outperforms RF and Linear (2020b)
bands) amphibolite, granite, acidic discriminant analysis (LDA) in lithological
intrusive and migmatite). mapping.
SAM and Sub- PRISMA Mafic and ultramafic rocks 3 – The ground control points validation Libeesh and
pixel (Pyroxenite, Peridotite, Gabbro). exhibits the high performance of sub- Arivazhagan
algorithms pixel algorithms in lithological (2023)
classification using PRISMA imagery.
M3D-DCNN, AHSI Hornstone, basalt, slate, limestone, 14 >90 3D multi-scale deep learning Ye et al. (2020)
SSUN, marble, granite, diorite, classifications show with hyperspectral
HybridSN, granodiorite. data outperform the conventional ML
SVM models in 3D lithological classification.

ISODATA and K- DESIS Mixed classes of Basalt and PRISMA: 7, 15; – The processing of hyperspectral data by Tripathi and Garg
means PRISMA Minerals DESIS: 3, 4, unsupervised learning and dimensionality (2023)
respectively. reduction techniques allowed the
mapping of lithological units. In
particular, K-means revealed relatively
more accurate results.
PCA-QPSO- TASI LiuYuan town, Guazhou, China 27 74.36 The novel PCA-QPSO-LSSVM Fang et al. (2022)
LSSVM lithologies including Andesite , combinational model has been
dolomite marble, marble, two-mica demonstrated as a practical classifier of
quartzite, monzonitic granite, hyperspectral thermal data, showing
monzodiorite, biotite plagioclase accurate lithological units mapping
granulite, granite porphyry. results than the conventional LSSVM.
SAM GaoFen-5 Metamorphic quartz sandstone, 3 86.75 SAM on GF-5 hyperspectral data for Cai et al. (2020)
metasandstone, pyllite complex lithological mapping in
mountainous terrain, providing insights
for regional investigations and decision-
making through accurate assessment.
EnGeoMAP and HyMap Dunite, peridotite, pyroxenite, 4 – According to the findings, ISMA is more Salehi et al.
ISMA gabbro effective when mapping rocks consisting (2020)
of peridotite and pyroxenite, while
EnGeoMAP performs better when
mapping the dunite unit.
(continued on next page)

10
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Table 3 (continued)

Method Dataset Feature Classes number OA Research Main Highlights Reference

(%)

SAM Hyperion Rocks/Minerals (Gypsum, Marl, 5 – SAM application to Hyperion data Thannoun (2021)
Sandstone, Jarosite, exhibits promising rocks/minerals
Montmorillonite) mapping results. Contributing in mineral
prospecting at the Low Folded Zone of
Iraq.
SAM, MTMF CASI/SASI 14 Rocks/Minerals including 14 81 The using of SAM, MNF, PCA and MTMF Wan et al. (2021)
granite, porphyry granite, to process high-resolution hyperspectral
phylliteschist, iron-manganese cap data allowed an effective mineral and
lead–zinc oxide ore body, skarn, geological mapping, aiming for mineral
marble exploration.
SAM, MTMF Specim Calcareous shale, silty limestone, 3 – Ground-based LiDAR and hyperspectral Shahtakhtinskiy
dual-camera pure limestone sensor data were used for point cloud- and Khan (2020)
system registered lithological mapping. The
resulted MTMF-based RGB false color
composite interpretation and SAM
classification offer detailed insights for
geological analysis.
MTMF SEBASS Quartz-rich sandstone, quartzite, – – TIR suboixel-based mapping effectively Aslett et al.
conglomerate, and alluvium; used in mapped very similar (2018)
muscovite-rich schist, siltstone, and metasedimentary rock units.
slate; and carbonate-rich dolomite,
limestone, and marble,
SAM Telops Lithologies including sedimentary , 11 TIR hyperspectral aerial imagery was Schodlok and
Hyper-CAM and metamorphic rocks, in addition conducted to map the lithology and the Frei (2020)
to mineralization endmember zinc mineralization.
HSS-LSTM AVIRIS Nevada minig area 8 94.70 HSS-LSTM is demonstrated as a Zhao et al.
hierarchical spatial-spectral feature (2020)
extraction method, achieving higher
hyperspectral lithological classification
performance compared to other methods.
Semi-supervised Hyperion Dolomite, shale and marl, siltstone 7 – SSL method using limited samples Guo et al. (2021)
self-learning and shale, quartzose sandstone, number has achieved accurate results in
(SSL) conglomerate, sandstones and lithological mapping than active learning
conglomerate. models implementing a complete samples
number.
SVM UAV 3 zones most rocks are magnatic, 4, 7and 7 – WSRRR exhibited the best results among Lorenz et al.
sedimentary, and metamorphic the tested methods, while OTVCA was (2021)
comparatively slow. Refining the method
could lead to increased processing speed,
a significant step towards (near-)real-time
data processing
RF AISAFenix 15 Rocks/Mineral/(Vegetation) 15 84% MNF transformed AISAFenix bands have (Kereszturi et al.,
including alteration minerals, been used to classify rock/alteration 2018)
sedimentary rocks, and volcanics minerals, further improvement
(OA=85%) have been obtained by
combining the HIS data with LIDAR
Sentinel-2A and OLI + TIR

Province, China. The maximum OA achieved by CNNs across these three sites was very high with values more than 94%. In the study
conducted by Ye et al. (2020), the great effectiveness and consistency of DL methodologies were examined by employing a spectral-
spatial unified network (SSUN), a hybrid spectral CNN (HybridSN), and a multi-scale 3D deep convolutional neural network (M3D-
DCNN). These DL models were evaluated in comparison to the SVM technique, utilizing six datasets established from GF-5 AHSI,
SASI, and Sentinel-2A imagery. Hajaj et al. (2023b) used CNN and HyMap data for lithologic mapping within the Ameln valley region
(Anti-Atlasic belt, Morocco), yielding an outstanding overall accuracy of approximately 95% (Fig. 5). The study underscores the po-
tency of combining CNN and HyMap data for optimal lithologic mapping in the study region. Nevertheless, the research underscores
the complexities inherent in distinguishing distinct lithological units through remote sensing data, citing challenges stemming from
shared chemical and mineralogical compositions that lead to spectral resemblances.
Qin et al. (2019) proposed a Semi-supervised learning (SSL) model called dilated-CNN for an optimized lithological classifica-
tion, using HSI, the dilated-CNN principle is based on increasing receptive field size without using max-pooling. Dilated CNN has
been developed for the first time for segmentation tasks, aggregating multi-scale contextual information while preserving resolution
(Yu and Koltun, 2015).
SSL classifications use the small amount of available labeled data with the unlabeled data to train classification. The SSL models
were recognized as a hybrid learning models, which incorporate the strengths of both unsupervised and supervised models. Genera-
tive Models (Jin et al., 2013), Co-training Models (Blum and Mitchell, 1998), Self-training Models (Dópido et al., 2013), Graph-based

11
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 5. Lithological mapping results in three zones from the Ameln valley shear zone, eastern Kerdous inlier (Moroccan Anti-Atlasic belt) using SVM-RBF, SVM-LK,
SVM-PL, KNN, RF, and 1D-CNN algorithms (Hajaj et al., 2023b).

methods (Camps-Valls et al., 2007), and Transductive Support Vector Machines (Bruzzone et al., 2006) are the main five SSL models
categories. As an example, Guo et al. (2021) employed a multinomial logistic regression (MLR) SSL approach in their study to en-
hance the accuracy of lithological mapping through the integration of both spectral and spatial data. The used self-training-based
model classification results have been evaluated comparatively with the results obtained using the RF-supervised model using the
training complete dataset. Then, experimental findings demonstrate that SSL approach achieves accurate results. In the realm of HSI
classification, a survey of SSL models has been conducted by Sawant and Prabukumar (2017).
Unsupervised algorithms show more interest but still have the non-requirement of defining targeted geological material spectra
advantage. These algorithms are appropriate for non-studied regions and zones with poor geological backgrounds. Recently, the
study by Tripathi and Garg (2023) showcased the ability of unsupervised classifiers and FE techniques in exploiting satellite hyper-
spectral remote sensing data for lithological classification in Banswara, Rajasthan, India. Unsupervised ML methods namely Iterative
Self-Organizing Data Analysis Technique (ISODATA) and K-means were employed for lithological mapping using PRISMA and DESIS.
The research utilized VCA to evaluate similarity and detect distinctive spectral characteristics. The research underscores the superior-

12
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

ity of KPCA and PCA over ICA and MNF in FE. Then, the results revealed that K-means emerges as the recommended method for hy-
perspectral-based lithological classification in the area.

4.4. Mineral prospectivity mapping

4.4.1. Mineral detection and mapping
Hyperspectral technology has been extensively employed in mineral mapping investigations. A variety of ore deposits can be ad-
dressed through hyperspectral data, encompassing supergene, hydrothermal, and magmatic mineral deposits (Bedini, 2017). Table 4
shows the main machine learning methods adopted in mineralogical identification and mapping.
The per-pixel classification generates and allocates a distinctive signature by gathering the feature spectra set across the entirety
of the pixel to a specified category. Without accounting for mixed pixel states, the resulting signature unites all materials present in
the pixel's training dataset (Pal et al., 2020). Besides, sub-pixel classification methods have been introduced to address the complica-
tion of multiple distinct materials coexisting within a pixel. These methods regard each pixel spectral value as a result of a linear or
non-linear pure materials combination. Such approaches accurately designate a specific category for every pixel, particularly applica-
ble to medium to low spatial-resolution hyperspectral imaging. Various linear and non-linear unmixing algorithms were employed to
accomplish sub-pixel level spectral alignment in this context (Lu and Weng, 2007; Pour et al., 2019).

4.4.1.1. Sub-pixel techniques. Wan et al. (2021) applied MTMF to classify hyperspectral alteration minerals in a polymetallic
Au–Ag–Pb–Zn ore deposit located in Huaniushan, Gansu Province. Their findings demonstrated the effectiveness of MTMF, comple-
mented by additional processing techniques such as MNF and SAM, in enhancing previous mineral and geologic maps. These iden-
tified targets represent significant areas warranting further investigation in mineral resource exploration, as shown in Fig. 6. Hajaj
et al. (2023a) used MTMF to map alteration minerals and mineral mixtures in Ameln valley shear zone (Anti-Atlas, Morocco) by

Table 4
Sub-pixel, per-pixel, ML, and DL methods applied to hyperspectral imagery for mineralogical identification and mapping.

Method Dataset Feature Exploitable Reference

element

MTMF ZY1-02D White mica, carbonate minerals Cu Wu et al. (2023)

HyMap Classifying 14 minerals and mineral mixtures Mn, Cu Hajaj et al. (2023a)
Hyperion Propylitic, sericitic, argillic, and FeOx (iron oxide) Base metals Canbaz et al. (2021)
alterations
CASI/SASI Kaolinite, calicite, sericite, chlorite, diopside, jarosite, iron Polymetallic Wan et al. (2021)
carbonite
CEM, MF AVIRIS Alteration mean spectra Au Guha et al. (2021a)
SAM AVIRIS-NG Kaolinite, talc, montmorillonite Base metals Mishra et al. (2022a)
AVIRIS-NG Goethite and limonite, kaolinite, talc, kaosmec, Base metals Mishra et al. (2021)
Hyperion High potential zones of iron mineral. Fe Sengupta et al. (2019)
Hyperion Alunite, axinite, illite, and pyrite, goethite and hematite Fe, Cu Govil et al. (2021)
Hyperion Goethite, hematite, kaolinite Fe Souza et al. (2021)
Hyperion 5 classes of hematite Fe Shaik et al. (2021)
AVIRIS-NG Chlorite, calcite, muscovite, epidote Base metals Samani et al. (2021)
SVM HyMap Mapping regolith-geology thematic map targeting Ni Ni De Boissieu et al. (2018)
mineralization associated with weathered peridotite
SFF, SAM, OSP and Hyperion + TIR(ASTER) Alunite, calcite, kaolinite, montmorillonite, muscovite, Au (Not known as Ni et al. (2020)
ACE gneiss, marble, quartz economic)
LSU, SAM, SFF and AVIRIS-NG Talc Talc Bhadra et al. (2020)
MTMF
SAM, and SFF AVIRIS-NG Montmorillonite, Talc, Dolomite Base metals (Bhadra et al., 2020;
Tripathi and Govil,
2019b)
SFF HyMap Alteration minerals including alunite, pyrophyllite and Cu–Au Caruso et al. (2021)
dickite
K-means AVIRIS-NG Kaolinite, montmorillonite, alunite, muscovite, calcite, Au (Not known as Ren et al. (2020)
chalcedony economic)
ML algorithms PRISMA Montmorillonite, talc, kaolinite Fe Agrawal et al. (2023)
RF Hyperion Dolomite Zn Aravinth et al. (2019)
NEO HySpex Goethite, hematite, jarosite, kaolinite, calcite, epidote, and Cu (Au, Mo) Wang et al. (2022)
chlorite
Decision Trees AisaFENIX (SWIR), Microcline, albite, plagioclase, cookeite, orthoclase, Li Booysen et al. (2022)
AisaOWL (LWIR) montebrasite, muscovite, ammonium-bearing minerals, and
quartz
Ensemble Learning ZY1-02D Limonite, chlorite Au-polymetallic Lin et al. (2022)
DL GoeFen-5 Porphyry copper-related Alteration minerals Cu Fu et al. (2021)
DL (ASMLP) AVIRIS-NG Kaolinite, montmorillonite, alunite, muscovite, calcite, Au (Not known as Deng et al. (2021)
chalcedony economic)
CNN AVIRIS-NG Chalcedony, Muscovite, Montmorillonite, Kaolinite, Fe Agrawal and Govil (2023)
Halloysite, Calcite, Alunite.
DNN SASI Mica, calcite, montmorillonite, chlorite U Zhang et al. (2022)

13
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 6. Extract of MTMF results, generated by Yu-Qing Wan, using ENVI (version 4.3) created by the ENVI Team under an open license (http://www.enviidl.com/)
(Wan et al., 2021).

processing HyMap data. Petrography, fieldwork, and XRD analysis were used in the validation of the mapping results of 14 target
minerals. New prospection data were derived essentially from the mineral mapping of the study area. As well as the HyMap im-
agery processing using MTMF allows performing a high-resolution mineral mapping of the study area around the Idikel mine. In
(Guha et al., 2021a), a synergistic approach involving AVIRIS-NG, ASTER, and L-band ALOS PALSAR data was employed to investi-
gate lode-type gold deposits in the Hutti Maski Schist Belt, India. Mineralogical and petrographical analysis of altered rocks in the
study area demonstrated the presence of calcite and scapolite. The representative reflectance alterations spectrum was utilized to
delineate the small, irregularly shaped alterations' spatial distribution. A consolidated map of altered rocks was generated by com-
bining two independent alteration maps produced through the constrained energy minimization and matched filtering sub-pixel
models. The identification of altered rock occurrences near favorable lineament and their closeness to contacts between metavol-
canic and younger felsic intrusive formations underscores their potential significance as sites warranting thorough exploration.
Canbaz et al. (2021) applied MTMF to Hyperion scene to delineate the alteration minerals within a target area from Sivas-Turkey.
The sub-pixel mapping targets hydrothermal alterations encompassing iron oxide (FeOx), propylitic, sericitic, and argillic alter-
ations. These alteration features have been targeted based on the altered sample surfaces analysis by the ASD FieldSpec instrument.
The results demonstrated a strong alignment between field and laboratory investigations, as well as the MTMF outcomes, yielding
an overall accuracy surpassing 86% with a kappa coefficient of 80%. Notably, the MTMF-based alteration mineral mapping marked
the region's preliminary hyperspectral exploration into the detailed mineralogy of hydrothermal features. Therefore, the mapping
outcomes aid as a reference for alteration assessments during hydrothermal base metal mineralization exploration.
Wu et al. (2023) employed the MTMF method to enhance the precision of extracting iron oxide alterations. For hyperspectral
data, pixel index algorithms were used to derive endmembers for white mica and carbonates. Furthermore, the alteration mapping
was done by MTMF method, this latter allows abundance mapping to intricately classify white carbonatite and mica alterations into
five distinct categories. The outcomes from multisource remotely sensed datasets processing were subsequently subjected to a com-
parative and analytical examination, designed to mitigate sporadic single-image extraction outcomes. This evaluation substantiated
the superior performance of hyperspectral remote sensing in conjunction with the MTMF approach. Reinforced by field verification
was used for validating the precision of the alteration information extraction in the Secuo, Mamupu, and Jicuo deposits.
MTMF revealed successful results in mapping the associated hydrothermal alteration to several ore deposits. Besides, Dong et al.
(2022) used the minerals absorption feature wavelength positions with MTMF technique to process GF-5 data for mineralogical map-
ping in a porphyry and epithermal ore district. Fig. 7 displays the resampled minerals spectra sourced from the USGS spectral library,
into the GF-HS dataset resolution. The porphyry Cu deposits alteration minerals spectral features area typically in the 2–2.4 μm spec-
tral range Fig. 7. A detailed mapping of porphyry-epithermal Cu–Au deposit associated alteration has been performed based on the al-
teration minerals’ absorption wavelength positions and the MTMF algorithm, where advanced argillic typical alteration minerals and
white mica varieties were depicted. A CASI/SASI imagery from Huaniushan Au–Ag–Pb–Zn ore district–Gansu has been processed us-
ing MTMF for mapping hydrothermal alteration minerals (Fan et al., 2021). Using the same HS sensor, MTMF model has been used in
mapping hydrothermal alterations in Baiyanghe Uranium (U) deposit, Xinjiang, NW China. Zoning features of kaolinite, montmoril-
lonite, chlorite, white mica, carbonate, and ferric iron were revealed in the area (Xu et al., 2018). As well, Epithermal deposits associ-
ated mineral alterations have been mapped using MTMF. CASI/SASI scene covering Wuyi metallogenic belt, China was used the map
abundances of alteration minerals and their distribution (Huang and Zheng, 2018). High sulfidation epithermal targets were investi-
gated in Chapi Chiara gold prospect, Peru using hyperspectral data. The HyMap scene of the area has been processed using MTMF to
map hydrothermal alteration minerals paragenesis of argillic, advanced argillic, and propylitic alteration (Carrino et al., 2018).

14
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 7. Pure minerals spectral signatures sourced from USGS spectral library resampled to the of the GF-HS data resolution (Dong et al., 2022).

Jain and Sharma (2019) study focuses on utilizing remote sensing technology for mineral mapping, employing algorithms,
namely SAM, SFF, and MTMF with airborne hyperspectral data. The study identifies 13 endmembers grouped into clay, iron, car-
bonate, and other mineral. MTMF algorithm offers more accurate results compared to other algorithms, providing detailed sur-
face mineralogical information crucial for geological assessments, moreover, it demonstrates the highest ability in minerals-
settlements distinction.
LSU sub-pixel model (Chang, 2003; Chang et al., 2014) has been applied to HS LWIR data to map igneous rocks and hydrothermal
zones from Naundorf quarry (Germany) in the study by Lorenz et al. (2018a). Quartz-calcite-albite minerals were successfully distin-
guched from their host rock. HS SU models have been detailed in Section 5, with examples.

4.4.1.2. Per-pixel techniques. Indeed, SAM has known a great interest and it is one of the widely used algorithms in mineral mapping
studies using HSI (Kayet et al., 2018; Kumar et al., 2018; McCormick et al., 2021). Mishra et al. (2022b) applied the SAM algorithm
to AVIRIS-NG to map alteration minerals in Jahazpur, Rajasthan (India), and several conventional and analytic approaches were
used for verification. Within the same study area, Mishra et al. (2022a) conducted a comparative analysis that underscores the capac-
ity of the AVIRIS-NG and PRISMA sensors to distinguish altered or weathered minerals using SAM algorithm. The classification ob-
tained from the PRISMA shows that the latter was unable to detect minor surface exposures of talc deposits at the Padampura site. In
contrast, the AVIRIS-NG imagery effectively delineated this exposure within the classified mineral map. This difference in mapping
has been attributed to the relatively low spatial resolution and SNR. Sengupta et al. (2019) utilized the N-FINDER method followed
by the application of the SAM algorithm to discern hematite and goethite exposures through spectral analysis. In addition, band ratio
technique has been applied to ASTER data and integrated to map ferric iron, ferrous silicate, ferric oxide, and ferrous iron mineral

15
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

distribution. The Kappa Coefficient exhibited a precise representation of iron ore alteration within the mapping process. Govil et al.
(2021) employed the SAM technique along with field surveys within two designated prospective mineralized regions. The outcomes
of these investigations validated the existence of gossan and altered minerals, previously identified via Hyperion hyperspectral data
using the SAM method. This study showcased the effectiveness of hyperspectral data in conjunction with on-site field examinations
for the recognition of hydrothermal alteration minerals, gossans, and potential mineralized regions in the Himalayan context. Awad
et al. (2018) used the Ground-measured spectra of kaolin with varying quality and structural order-disorder to build an input of SAM.
As a result, SAM classifier has been successfully employed to map Kaolin distribution based on Hyperion data.
Wang et al. (2021b) proposed a wavelet SAM (WSAM) tri-training model to map alteration minerals at Dehua-Youxi-Yongtai Ore
District using an airborne HSI. Accuracy assessment showed an OA of 82% and 70% for virtual and XRD-based approaches, respec-
tively. Applied SAM to HSI, to detect Skarn ores outcrops based on their field spectra or based on associated alterations. The results
demonstrated the possibility of targeting Skarns directly from their spectra using HSI rather than mineral proxies. SAM classification
applied to HSI identified the Skarn potential zones, faciliting mineral prospecting in Xinjiang Uygur autonomous region, China.
Shaik et al. (2021) proposed a methodology for the mapping hematite ore categories based on their absorption spectra across the
Bailadila hill range in the Chhattisgarh State of India. Hyperion data was processed by SAM algorithm, then, spectral feature fitting
(SFF) technique was used for validation. The findings of the study demonstrated a noteworthy association between the concentration
of iron oxide and the attributes of the near-infrared absorption feature, as indicated by the correlation coefficients (R2) of 0.843 for
depth and 0.812 for width. This relationship was established through the application of various empirical models, with root-mean-
square errors (RMSE). Samani et al. (2021) used AVIRIS-NG and SAM algorithm for evaluating and mapping possible mineral alter-
ations in the Ambaji–Deri Region (Northwestern India). The USGS spectral library was used as a reference to study the matching of
the AVIRIS-NG-derived spectra. Thereafter, ground geological records were used in verifying classification findings. Globally, good
mapping results were observed, except some difficulties in differentiating chlorite and epidote. To handle this limitation, additional
chemical data are proposed by the author. Souza et al. (2021) used the SAM to process Hyperion for regolith mapping in Midwest,
Brazil. Employing reflectance spectroscopy at both regional (using OLI) and local (using Hyperion) levels, substantiated by on-site
verification, demonstrates the high effectiveness of these sensors when combined with mathematical algorithms. This alliance facili-
tates the cartography of the spectral attributes of the tropical regolith developed under conditions of moderate to robust lateritiza-
tion in the Midwest region.
De Boissieu et al. (2018) applied SVM algorithm to generate the regolith-geology classification map of the New Caledonia, aiming
for Ni exploration. The accuracy of this map was evaluated in comparison with field sites where regolith had been interpreted, re-
sulting in an approximate 70% congruence across all unit types. Additionally, a comparison was made with a regolith-geology map
produced through conventional data sources, not utilizing HSI. Crucially, the mineral map derived from hyperspectral data offered a
significantly higher level of detail, facilitating a more refined comprehension of the regolith-geological composition, particularly in
areas with exposed soils and rocks. Ni et al. (2020) investigated the integration of the multispectral TIR dataset to Hyperspectral
VNIR/SWIR dataset for lithological mapping using the SAM, spectral feature fitting (SFF), the adaptive coherence/cosine estimator
(ACE), and orthogonal subspace projection (OSP). The ACE classification achieved the highest classification accuracy, followed
closely by the OSP classification. The SAM and SFF classifiers have slightly lower accuracies. Even with a low spectral resolution, the
TIR dataset remains an efficient complement to hyperspectral VNIR/SWIR dataset for mineral mapping tasks. Ren et al. (2020) con-
ducted a case study in Cuprite, Nevada, using an improved k-means algorithm that exhibits a notable capability in identifying the
majority of minerals, achieving a kappa value exceeding 0.8. This value is significantly higher by 46% and 15% in comparison to the
conventional k-means and spectral matching techniques, respectively. The introduction of new mineral categories becomes more
probable as the value of K increases. Notably, when K greatly surpasses the count of distinct mineral categories, the precision of re-
sults improves and the mineral mapping outcomes become less dependent on the methodology used for similarity measurement. Fur-
thermore, the improved k-means algorithm effectively diminishes speckle noise from mineral mapping outcomes and holds the po-
tential to identify other objects as well.
Rialland et al. (2021) used a fuzzy logic model to identify minerals from reflectance spectra extracted from hyperspectral sensors
operating in the VNIR and SWIR ranges. This technique is based on the fuzzy set theory (Zadeh, 1965), it was applied to analyze data
collected from airborne HySpex and satellite-borne PRISMA images during a specific campaign conducted over two quarries in
France. The fuzzy logic system utilized in this study mimics human reasoning, showcasing its applicability for automated mineral
identification across various mineralogical scenarios, including situations involving mineral mixtures.
In the research conducted by Caruso et al. (2021), the SFF and XRD-validation using facilitated the interpretation of mineralogical
variations. This investigation highlights the feasibility of detecting surface alterations linked to potential mineralization through the
application of airborne HSI and semi-quantitative XRD, even within the regions predominantly characterized by weathered regolith.
Machine learning algorithms have a gained more interest in mineral detection and mapping nowadays. Booysen et al. (2022) em-
ployed a decision tree model with a multi-scale hyperspectral imaging strategy to delineate lithium-bearing minerals and their respec-
tive spatial patterns within pegmatite formations across three distinct scales. This endeavor facilitated the creation of an accurately
aligned and seamlessly connected 3D mineral map for the principal excavation site. Notably, the study highlighted the significance of
precise processing within a 3D framework and highlighted our capacity to directly chart lithium-bearing minerals within outcropping
rock formations by hyperspectral imaging. Following the combining of multiple Decision Trees to create a more robust and accurate
prediction, RF are ensemble learning algorithm that uses the voting operation to classify the target in the image (Breiman, 2001). In a
study conducted by Wang et al. (2022), Random Forests were employed to process a NEO HySpex imagery encompassing porphyry
Cu (Au, Mo) mineralization in China's Kalatag District. The research effectively discerned and mapped alteration mineral assemblages
linked to iron oxide/hydroxide, clay, and propylitic alterations (see Fig. 8). The utilization of Random Forests yielded remarkable

16
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 8. Alteration mineral map derived from NEO HySpex imagery for Yudai, China's Kalatag District (Wang et al., 2022).

classification accuracy (73.08%) and Kappa classification values (65.73%) for identifying alteration minerals. Lin et al. (2022) inte-
grated the Sparrow Search Algorithm (SSA) with two ML models, namely the RF and Gradient Boosting Decision Tree (GBDT) ensem-
ble model to map minerals from ZY1-02D hyperspectral data within the Qinghai Gouli area, China. The evaluation of these models' ef-
ficacy in mineral information extraction was conducted using metrics such as the Youden index (YD) and ore deposit coincidence
(ODC). The findings highlight an enhancement in the optimization of the SSA algorithm parameters. Moreover, the ensemble models'
accuracy has significantly increased after a parameter search. Notably, the SSA-GBDT model demonstrated the most optimal perfor-
mance in comparison with conventional machine learning approaches.
Aravinth et al. (2019) used RF to classify Zincian Dolomite based on Hyperion data, as a results, RF reveals the highest OA com-
pared to K-means, and SVM. The suggested method for mineral exploration is proven to be more accurate in identifying minerals
than conventional methods while also being faster and less expensive.
In the realm of Drill-core mapping, Acosta et al. (2019) introduced a novel approach that integrates detailed mineralogical analy-
sis with hyperspectral (HS) data. This requires creating training labels using an automated high-resolution mineralogical imaging
technology, such as mineral liberation analysis by scanning electron microscopy. After that, the SEM-MLA pictures are resampled to
match the HS data's resolution, and mineral mapping is done using a soft labeling technique. Labels presented geological mixtures of
interest and utilize classifiers i.e., SVM, and RF to map the entire drill-core. The framework was validated both qualitatively and
quantitatively, demonstrating its effectiveness in merging and enhancing high-resolution mineralogical analysis with drill-core HS
data. Contreras et al. (2019) utilized extreme learning machines (ELM) for mineral mapping in drill core hyperspectral data. Princi-
pally, ELM model aims to reach the lowest training error with the smallest weights norm (Huang et al., 2006). To implement this tech-
nique, a reference map was derived from the drill core hyperspectral dataset. The findings demonstrate the ELM usefulness for HS
minerals mapping in drill core, yielding superior quantitative and qualitative outcomes compared to the RF algorithm.
Over the past few years, the paradigm of DL in computing has brought about a transformative shift in the approach to analyzing
remote sensing data (Li et al., 2023; Xu et al., 2021; Zhang et al., 2022). Agrawal and Govil (2023) evaluated a PRISMA scene cover-
ing Jahazpur in Rajasthan (India) following the use of various ML and DL classifiers. To perform classification, a SAM-derived labeled
data showing the distribution of talc, kaolinite, and montmorillonite distributions has been implemented. Field surveys in the ex-
posed altered outcrop were conducted to validate the accuracy of the mineral distribution. Multiple metrics (e.g., OA, KC, F1-score,
area under ROC), were used to evaluate the classification results of ML and DL techniques. The study highlighted the effectiveness of
classifiers employing stochastic gradient descent and multilayer perceptron (MLP), showcasing the significant potential of PRISMA
data for alteration minerals mapping.
Fu et al. (2021) used DL models to the GF-5 hyperspectral dataset of the Duolong district (porphyry-epithermal system) to inves-
tigate their utility in hydrothermal alteration mapping. Stacked autoencoder and various computationally efficient convolutional
CNN were evaluated for their potential in extensive mineral exploration. Notably, the mixed convolutions and covariance pooling
technique yielded the highest classification accuracy (>96%). The feature fusion approach, due to its lightweight design, was also
endorsed. These two spectral-spatial classifiers exhibited strong classification performance by mitigating misclassification between
similar minerals and reducing noise, showcasing their robust learning capabilities.
Zhang et al. (2022) used three deep neural network models to process the SASI hyperspectral scene covering the Baiyanghe ura-
nium deposit, Northwestern Xinjiang, China. The study centered on creating and contrasting three deep neural network (DNN) mod-
els, namely 1D-CNN, a fully connected neural network (FCNN), and a hybrid CNN incorporating both 1D and 2D-CNN for mineral
mapping. The evaluation of these models yielded the best classification results by using 1D and 2D CNN combination. These DNN

17
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

models were applied for mineral identification and mapping within the SASI hyperspectral scenes in the Baiyanghe uranium mining
region. The accuracy of the 1D and 2D CNN outperformed other methods, especially for distinguishing minerals with analogous
spectral traits like short-wavelength white mica and medium-wavelength white mica, highlighting the effectiveness of spatial infor-
mation incorporation in enhancing mineral identification in hyperspectral imaging, as well as pointing the CNN higher accuracy.
Deng et al. (2021) proposed a mineral classification methodology using the Attenuation Spectral Absorption Index (ASAI) to de-
tect mineral's spectra intrinsic features from combining spectral absorption index (SAI) and absorption position, then, applying the
multilayer perceptron (MLP) multilayer feed-forward neural network. As result, the model achieved 93.62% overall accuracy in
mineral identification using AVIRIS hyperspectral data covering the Nevada mining area.
In drill-core mineralogical mapping, Zhang et al. (2023) propose a novel approach denoted graph convolutional neural networks-
SAM (GCNNSAM). This approach combines the strengths of DL and spectral matching to identify mineral data from hyperspectral im-
ages of drill cores. Comparing the performance of GCNN, SAM, and GCNNSAM mapping techniques. GCNNSAM achieves the highest
OA (89.23%) in mineralogical classification.

4.4.2. Prospectivity mapping

The elaboration accurate mineral potential map can require the integration of various data types (e.g., geophysical, geochemical,
structural, lithological, and spectral). Using Geographic Information Systems (GIS), experts can time effectively assess spatial geo-
science data to map zones of high mineral potentiality for future exploration processes. The process of quantitatively assessing geo-
logical information for integration into mineral potential mapping involves various techniques, such as neural networks (Fung and
Wong, 2005), fuzzy logic (Bonham-Carter, 1994), evidential belief (Carranza and Hale, 2003), and weights of evidence (Agterberg et
al., 1993). By the using of hyperspectral data the mineral detection can be more precise. Thus, the hyperspectral-derived layers will
contribute well in the consolidated mineral potential map. Ferrier et al. (2019) conducted a mineral prospectivity mapping (MPM) of
the epithermal deposits on Milos Island, Greece. They used GIS and a fuzzy multi-criteria evaluation approach to analyze the spatial
distribution of deposits, as well as the study aimed to understand factors influencing ore deposition. A combination has been carried
out from based on DAIS HSI, multispectral thermal data, and field data to generate MPM of the area. The final model incorporated in-
formation on silicified lithocaps, altered rocks, faults, and topography to provide insights into mineral potential. In the Hajaj et al.
(2023a) study, airborne hyperspectral data was employed to perform a high-resolution mineral potential map for eastern Kerdous in-
lier (Moroccan Anti-Atlas). This approach utilized fuzzy modeling to integrate various thematic layers, including RBD, Crosta tech-
nique, MTMF, and Line density, derived from HyMap imagery. The MPM highlighted three new potential zones for copper and man-
ganese exploration in the study area. In another study, Mondal et al. (2022) combined AVIRIS-NG data and ground geophysical anom-
alies data to detect chromite-bearing mafic cumulates in the Sittampundi Layered-Complex, Tamil Nadu, India. The Matched Filtering
(MF) algorithm was applied to a specific band set of AVIRIS-NG data, generating a spectral map indicating chromite-bearing mafic cu-
mulates, thereafter, a low pass moving kernel filtering has been applied to the resulted spectral map. Then, SVM was used to integrate
spectral maps, magnetic anomaly, and residual gravity maps. For validating the identified chromitite occurrences distributions, geo-
logical maps and field data have been used, revealing the effectiveness of the HSI and geophysical data combination.
Fu et al. (2023) proposed CNN model for large-scale mineral prospectivity mapping porphyry copper deposits (PCDs) in the Duo-
long ore district, Tibet, China. The mineral prospectivity mapping was based on hyperspectral imagery supplemented by multispec-
tral imagery and geochemical element data. CNN achieves the greatest classification performance with an accuracy of 0.956. Four
metallogenic prospective sites were delineated and verified through field reconnaissance, indicating that the application of deep
learning-based methods in PCD prospecting is feasible by utilizing geoscience big data (RS and geochemical data).
A mineral prospectivity mapping study using by combining biogeochemistry, geology and HSI data was conducted in South Island
(Te Waipounamu in Te Reo) by Chakraborty et al. (2022). With an emphasis on pathfinder elements to gold-mineralization, many sta-
tistical techniques have been used to identify correlations between the biogeochemistry of the targeted plant and the underlying soil
and rock substrate. Furthermore, as highlighted in the research by Shirazy et al. (2020), statistical techniques yield promising out-
comes in delineating Ti–Zn mineralizations following a case study within the Kivi area, East Azerbaijan Province, Iran.

5. Research trends
5.1. Selection of optimum bands
In the realm of HSI, which comprises a multitude of closely correlated spectral bands, the meticulous selection of informative and
distinctive bands holds paramount significance for the successful classification of hyperspectral images (Sawant et al., 2021). Hyper-
spectral classification accuracy can be declined due to spectral information redundancy and noise, these issues could be handled by
the selection of the bands providing the highest level of information in the HSI (Harris et al., 2006). According to Sun and Du (2019)
hyperspectral band selection methods can be categorized to six categorizations, including embedding learning-based, sparsity-based,
clustering-based, ranking-based and searching-based methods.
Tan et al. (2020) introduced a novel approach for the selection of spectral bands in lithological mapping, wherein a criterion is
acquired through the assessment of lithological super-pixel uniformity and variance. The identification of discriminative bands per-
tinent to lithological distinctions is facilitated by a proficient clustering algorithm grounded in affinity propagation. Within the con-
text of this manuscript, Hyperion imagery from Junggar (China), featuring five distinct lithologic units, has been selected for vali-
dation. Fig. 9 displays the results of classification (using SVM classifier) based on the proposed BS method, all Hyperion bands,
VNIR bands, SWIR bands, and four other BS methods. Accuracy assessment results demonstrated that the best classification has
been derived from the proposed BS method, using few numbers of bands. Kumar et al. (2020b) introduced an innovative approach

18
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Fig. 9. Lithologic discrimination and geological mapping images of the Junggar. (a) Baseline with the all Hyperion 185 bands; (b) VNIR spectral channels; (c) Partial
SWIR spectral channels; (d) EDAP; (e) AAP; (f) E-FDPC; (g) SBS; (h) Proposed LSBS (Tan et al., 2020). Abbreviations: AAP = adaptive affinity propagation; EDAP=
Euclidian distance affinity propagation; E-FDPC = enhanced fast density-peak-based clustering; LSBS = lithologic superpixel band selection; SBS = superpixel-based
BS.

based on Joint Mutual Information Maximization (JMIM), demonstrating its superiority in selecting optimal spectral bands com-
pared to alternative input datasets. The application of these selected bands proved highly effective in accurately classifying a di-
verse range of lithological units, including magmatic rocks. Furthermore, a comprehensive sensitivity analysis undertaken in the
study revealed that SVM exhibited a notably lower sensitivity to factors such as sample size variations and mislabeling during
model training, as compared to other Machine Learning Algorithms (MLAs).
Lin et al. (2023) proposed an improved model based on the XGBoost framework combined with a greedy algorithm to accurate
hyperspectral lithological classification. This high-precision learning strategy consist on involving a greedy search algorithm at the
first layer of the two-layer XGBoost model, which is advantageous over the conventional optimal band selection techniques in the
analysis experimentations. To enhance the distinguishability between Li-rich pegmatite and Li-poor pegmatite using GF-5 AHSI data,
the optimal band combination was determined using both the least angular regression (LARS) and the genetic algorithm (GA) (Chen
et al., 2023). The minimum angle-fitting algorithm identified the top 20 bands that yielded the highest separability.

5.2. HSI unmixing

Spectral mixture is a rigorous challenge in target detection using HSI data. Numerous Spectral unmixing (SU) algorithms have
been proposed to solve this issue (Bioucas-Dias et al., 2012). The linear spectral unmixing (LSU) and non-linear spectral unmixing
(NLSU) models could be considered as the most known SU models categories. Whereas, the linear hyperspectral unmixing (LSU) mod-
els are more used by the authors due to their simplicity and the non-consideration of the scattering effects between endmembers
(Singer and McCord, 1979; Wei and Wang, 2020). Geometric, non-negative matrix factorization (NMF), Bayesian method, and sparse
regression are the main four categories of LSU (Feng et al., 2022; Wei and Wang, 2020). Another categorization devises SU model into
one- and two-step models, based on the identification of endmembers and their corresponding abundances, either in one step or in
two separated steps. In Hyperspectral mineral unmixing and mapping, several studies leveraged the usefulness of SU models (Combe
et al., 2008; Koirala et al., 2019; Schmidt et al., 2014; Siebels et al., 2020).
Some spectral unmixing (SU) models have been developed and applied to the newly launched PRISMA satellite data. For instance,
mineral detection outcomes in the Gara-Djebilet region of Southwestern Algeria, an NMF has been implemented in developing a mul-
tistage informed SU by Benhalouche et al. (2022). This two-stage algorithm involves iteratively detecting and removing spectrally-
known irrelevant minerals before the application of SU to the HSI.
As a semi-supervised LSU models for mineral abundance mapping, sparse unmixing can effectively prevent the drawbacks of ob-
taining incorrect endmember information and incorrectly estimating the endmember number in the HSI. Several cutting-edge sparse
unmixing algorithms that incorporate spectral-spatial information from HS imageries are presented in the study by Chen et al. (2024),
the latter demonstrated the effectiveness of these SU algorithms (e.g., SUnSAL: variable splitting and augmented lagrangian, SUnSAL-
TV algorithm by introducing a Total Variation (TV), SRSSWSU: spectral reweighting and spectral similarity weighting for sparse un-
mixing algorithm, etc.), on both simulated and real hyperspectral datasets. Various existing state-of-the-art sparse SU techniques were
reported in the study by (Chen et al., 2024).
In recent years, many DL-based techniques have been proposed for HS unmixing (Tao et al., 2022; Yang et al., 2022). These algo-
rithms have been categorized into autoencoder networks, convolutional neural networks, and convolutional autoencoder networks
(Tao et al., 2022). Indeed, the benefits of deep learning are evident in its capability to automatically identify features, and reduce ef-
fort by minimizing the need for human computation and feature construction. Furthermore, DL models offer a high level of accuracy,
and this accuracy can be further improved by increasing the volume of input data.

19
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

5.3. Cloud computing

In the realm of geological research, the concept of cloud computing has emerged as a pivotal element. Cloud computing, is charac-
terized as a form of virtual resources that autonomously manage and uphold exploration (Liu, 2020). It provides a valuable avenue for
professionals to seamlessly integrate, share, and analyze extensive sets of geological exploration data. The geological cloud-
computing platform was intentionally devised to cater to the growing demand for a network of services encompassing vast geological
data. This approach augments the realms of virtual and integrated resource technology, as well as remote user engagement (Liu,
2020). In the modern pursuit of identifying novel mineral deposits, innovative techniques have arisen, prompting the inception of Ge-
ological Cloud 1.0 in 2017 by Geological Survey China. An overarching challenge for geologists remains the constraint of siloed con-
trol over geological exploration databases, emphasizing the need to disseminate accurate geological insights. Cloud computing
emerges as a solution to this quandary (Liu, 2020), with the geological cloud aimed at sharing data resources through the automation
of basin exploration and production management. This is executed through the utilization of the VMware V sphere virtualization plat-
form to optimize existing hardware resources. The implementation of cloud computing for geological purposes encompasses Mi-
crosoft Azure and SQL Azure, alongside similar Microsoft cloud computing products, collectively underscoring the intricate interplay
between geological exploration data and their application within the geological cloud computing domain (Liu, 2020).
Addressing the simultaneous storage and processing of voluminous remote sensing data presents a considerable challenge (Zheng
et al., 2018). Propitiously, the emergence of cloud computing has introduced a promising solution. Cloud computing holds significant
potential in the realm of HSI, providing an efficient resolution for the storage and processing of HSIs. This is attributed to its capacity
for scalable storage and high-performance computing prowess (Zheng et al., 2021; Zinno et al., 2015). In the study conducted by
Ayma Quirita et al. (2022), a new distributed modified N-FINDR endmember extraction algorithm was introduced, designed to effec-
tively process extensive hyperspectral data volumes by exploiting the resources of computer clusters. The execution of this distributed
algorithm involved the expansion of the InterCloud Data Mining Package, initially designed for land cover classification, using the
HyperCloud-RS framework. This framework was adapted for endmember extraction, enabling its execution within cloud computing
environments. This approach grants users the capability to efficiently manage processing power and storage capacity to handle large
datasets. The empirical assessment analysis concentrated on both accuracy and execution time, examining various synthetic versions
of the AVIRIS Cuprite hyperspectral dataset. This investigation effectively addresses the challenges associated with the manipulation
of substantial hyperspectral datasets. It is important to note that the concurrent processing of advanced AI models (DNN as example)
by employing Field-Programmable Gate Arrays (FPGAs) and Graphical Processing Units (GPUs) stands out as an innovative strategy
to enhance computational efficiency without compromising model effectiveness (HajiRassouliha et al., 2018).

5.4. Data integration

The integration of geospatial data from multiple sources is proposed as a cost-effective and efficient strategy for conducting com-
prehensive analyses of diverse geospatial information, particularly in regions where geological data is limited (Chattoraj et al., 2020;
Ghamisi et al., 2019; Lorenz et al., 2018b; Wang et al., 2021d). Lithological and mineral mapping have known a notable improvement
using innovative approaches based on integrated multi-sources geoscientific datasets (Abedini et al., 2023; Kuras et al., 2022; Lorenz
et al., 2018b; Salehi et al., 2018; Shirazy et al., 2020; Wang et al., 2021c). Qasim et al. (2022) employed satellite multispectral data
along with laboratory HSI data for lithological mapping in the Naweoba block of Zhob ophiolite, western Pakistan. Hyperspectral-
based laboratory analyses the rock samples’ spectral signatures, alongside petrographic assessments, bolster the accuracy of lithologi-
cal mapping based on satellite imagery. The resulting lithological map demonstrates alignment with existing geological maps, then
the mineralogical analyses based on laboratory Hyperspectral SWIR camera prove their effectiveness and less time-consuming than
the conventional methods using thin sections.
Kuras et al. (2022) used HSI and aeromagnetic data fusion as a cost-effective and promising approach for enhancing primary
analysis of mineral deposits having a high iron-bearing content. The authors propose methods for integrating iron diagnostic infor-
mation from hyperspectral and magnetic data, allowing discrimination of different rock types based on their magnetic susceptibil-
ity. By empirically modeling statistical relationships, the study establishes connections between spectral absorption features and
magnetic responses from iron-bearing rocks. The integration process includes magnetic forward and magnetic inversion ap-
proaches. Jain et al. (2022) integrated a derived mineral map from the AVIRIS-NG dataset using MTMF algorithm with geological,
geochemical, and other RS multisource data in order to delineate lead–zinc bearing deposits within the Zawar mineralized belt, In-
dia. Promising results for mineral prospection have been obtained, indicated the existence of dolomites-mafics association.
Qin et al. (2021) coupled GF-5 alteration minerals data with Au-anomalies data to map Gold potential area in Chahuazhai Gold
Mining Area, China. The Regression models of Geographically Weighted Regression (GWR) a Back Propagation (BP) network, Par-
tial Least Squares (PLS), and regression Multiple Linear Regression (MLR) were implemented for the integration of the datasets.
Where GWR provides more reliable results than the other models. According to this recent advancements in HLM-MP data integra-
tion, the limited interdisciplinary studies amount (Van der Meer et al., 2012), seems to be more extended. However, the multidisci-
plinary in these studies is still very weak, for instance, the study of the relation between earth surface mineralogy/geology with
environmental quality, erosion models, health, or/and biology. The integration of UAS-borne HSI with magnetic data allowed a
straightforward mapping of ore mineralizations in the Otanmäki Fe–Ti–V deposit in central Finland (Jackisch et al., 2019), where,
MNF, SAM, and K-means were the main processing techniques applied to detect ore mineralization using UAS-borne HSI data.
In other studies aiming for precise 3D mineral mapping, the authors integrate photogrammetry data with HSI data, which can be
more beneficial in arctic and rugged topography regions (Lorenz et al., 2018b; Salehi et al., 2018, 2020). Lorenz et al. (2018b) high-
light the important role of ground-based hyperspectral imaging in geological mapping, particularly for challenging terrains such as

20
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

near-vertical cliffs and distant targets like mountain ridges. Aiming to achieve precise mineral mapping, a workflow with straightfor-
ward atmospheric and topographic corrections has been introduced in their research. The resultant spectral mapping products are
seamlessly integrated with 3D photogrammetric data to create geometrically calibrated hyperspectral data "hyperclouds" enabling ac-
curate representations of large-scale hyperspectral datacubes. In remote and challenging environments like the Arctic, Salehi et al.
(2018) propose an innovative approach that merges photogrammetry with vessel-based hyperspectral imaging data. This combina-
tion aims to provide quantitative mineral mapping data for geological outcrop models, facilitating differentiation between barren
lithology and mineralization. Fig. 10 presents a mineral classification results matched with a 2D pseudo-orthophoto. The study work-
flow involves corrective measures to eliminate distortion effects from vessel movements, spectral-based lithological mapping, and
precise integration of spectral products with photogrammetric terrain data. Experimental data from Greenland are employed to vali-
date the approach's accuracy. Horizontal viewing geometry, multi-source, and multi-scale data fusion approaches demonstrate a good
capability for regional mineral mapping.
Kurz et al. (2022) used close-range hyperspectral imaging to analyze Alacón Member outcrops in northeastern Spain, identifying
limestone–dolomite lithologies and quantifying dolomitization levels. The integration of photogrammetric and hyperspectral data,
as well, the lab analyses ensured empirical validation, revealing hyperspectral imaging's usefulness in identifying dolomite bodies
and correcting previous overestimations based on visual inspection.

6. Discussion: Synthesis and perspectives

The prior sections highlight the substantial advancements made in hyperspectral remote sensing techniques for both mineralogi-
cal-lithological mapping as well as mineral prospectivity over the last few years. The integration of hyperspectral datasets, from satel-
lite, airborne, UAS, and ground-based sensors, has significantly supported geological mapping studies. The rapid increase in remote
sensing data collection sources has stimulated the advancement of advanced data processing approaches. The significant importance
of ML and DL algorithms has not been highlighted enough in geoscientific applications. These algorithms have proven their effective-
ness in handling complex hyperspectral datasets, extracting valuable information, and notably improving classification accuracy.
The last decade presented the development of new hyperspectral sensors with wide spatial coverage. Accordingly, the challenges
associated with limited ground coverage in hyperspectral data, as extensively reviewed and addressed (Peyghambari and Zhang,
2021; Zhizhong et al., 2012), appear to be mitigated by the global coverage offered by the latter. Recent preliminary works have
demonstrated the promising results of PRISMA in lithological and mineral mapping (Agrawal et al., 2023; Bedini and Chen, 2020;
Bedini and Chen, 2022; Esmaeilia et al., 2024; Laukamp, 2022; Libeesh and Arivazhagan, 2023; Mishra et al., 2022a; Shebl et al.,
2023; Tripathi and Garg, 2023). However, the emerging HSI satellite data as PRISMA, HISUI, and EnMAP show a low ground sample

Fig. 10. 2D RGB pseudo-orthophotos with the resulting classification images from the: (a–c) Karrat and (d–f) Søndre Strømfjord, Arctic regions (draped on top).
Pseudo-orthophoto textured with MNF image, where bands 2, 5, and 7 are visualized in RGB (b,e); and pseudo-orthophoto textured with SAM classification results (c,f)
(Salehi et al., 2018).

21
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

distance (GSD) (30 m) which can be challenging for precise mineral/geological mapping. Appropriately, PRISMA developers tried to
handle this limitation by the addition of a panchromatic sensor (5 m) (Loizzo et al., 2016). Several studies have experienced difficul-
ties in reaching their intended outcomes due to the HSI low GSD, when compared to the surface exposure of the specific minerals/
lithologies under investigation (Bedini, 2017). HSI Super-resolution could be helpful to mitigate the derived errors from coarse spatial
resolution in HLM-MP investigations. Many of these techniques have been reviewed recently by Aburaed et al. (2023).
UASs could play an important role in terms of high-resolution HLM-MP. Although UAS-hyperspectral data has been used in a large
number of environmental and agricultural researches (Banerjee et al., 2020; Sankey et al., 2018; Wang et al., 2021a), a low number of
works for geoscientific aims have been published. However, the usage of UAS for mineral exploration has increased (Jackisch et al.,
2019; Malehmir et al., 2017). UAS HSI shows effective results in supporting mining campaigns in prospecting REE (Booysen et al.,
2020).
In recent studies using HSI data, various analytical techniques were used for analyzing and mapping mineralogical-lithological
features, including spectral dimensionality reduction, per- and sub-pixel methods, ML, and DL. In the HSI field, high dimensionality
poses challenges in processing and interpretation. Indeed, dimensionality reduction using unsupervised techniques including, ICA,
PCA, and MNF have emerged as effective solutions (Wang and Chang, 2006). These techniques streamline data processing, enhancing
classification accuracy and minimizing noise interference. Furthermore, dimensionality reduction techniques can be used with vari-
ous AI classification models to extract informative bands from hyperspectral data.
DNNs have proven to be a high-performance model in enhancing the detection of complex geological features in HSI datasets.
DNNs have been applied to hyperspectral data for tasks such as lithological classification and mineral detection, leading to increasing
the precision of the image processing results (Liu et al., 2021; Zhang et al., 2022). Indeed, a limited number of studies have been pub-
lished using DL techniques for lithological and mineral mapping. A wide range of DL algorithms are still without use for geoscientific
aims. The lack of the DL-based HLM-MP studies in a global scale will be more and more mitigated due to the recently launched plat-
forms. Various DL classifiers of HSI have been reviewed by Paoletti et al. (2019). The synergy of such AI methodologies with hyper-
spectral data is advantageous following their capacity to drive accurate and efficient interpretation of complex geological materials
(Yu et al., 2021). Advanced analytical algorithms including DL as well as multi-sources data fusion are recommended in hyperspectral
geological-mineralogical mapping in order to achieve accurate information about earth's surface. It is important to note that HS-
LIDAR data fusion still less exploited in HSLM-MP during the last years, while, some previous research were highlighted its benefit for
surface geology mapping (Buckley et al., 2013; Kurz et al., 2009). Furthermore, the application of DL-based models using HSI-derived
layers is still also very poor in mineral potential mapping (MPM) (e.g., Lin et al., 2021).
However, cloud computing studies for hyperspectral mineralogical-geological mapping are still limited. The use of cloud comput-
ing in hyperspectral geology will be very valuable in terms of handling the large volume of hyperspectral datasets, as well as provid-
ing more flexibility in their integration and sharing. The authors widely use knowledge-driven methods in this task, while the power
DL models could provide a high ability to extract high prospectivity area information. In this context, HSI data classification using DL
models has shown several challenges including, model complexity, long computational time, and the requiring a sufficient amount of
data in the ground truth sampling (labeled data).
In order to generate a new set of samples for HSI classification, generative adversarial networks (GANs), synthetic data generation,
and data augmentation techniques have been proposed (Wambugu et al., 2021). Additionally, other techniques could deal with in-
complete data, as is the case for Maximum entropy (MaxEnt). The general-purpose model of MaxEnt could use incomplete data for es-
timating the likelihood of target occurrence or non-occurrence by approximating the probability distribution that maximizes entropy
or closely aligns with a uniform distribution (Wu, 2012). This is achieved through incorporating constraints derived from the avail-
able incomplete dataset (Elith et al., 2011). The application of MaxEnt in mineral potential mapping demonstrated successful results
for both geochemical data (Zhang et al., 2019) and multispectral data (Shevyrev and Carranza, 2022). Additionally, various models
have been developed for mineral prospectivity mapping using MSI data including, Neuro-Fuzzy-AHP (NFAHP) (Shirazi et al., 2022)
and hybrid-Fuzzy-AHP (Khosravi et al., 2021), and they can be proposed for processing HSI data.
Advancements in HSI classification methods suggest a shift in the common pixel-based models to object based models, there's been
a shift towards a focus on identifying and analyzing image objects (Blaschke et al., 2014). In object-based models, HSI is segmented
into homogeneous regions (i.e., objects) by concatenating consecutive pixels with reasonably identical spectral and spatial properties.
Object detection models have been classified into categories, namely the supervised (based on prior knowledge of the object to be de-
tected) and the salient (Detecting and segmenting salient objects). Because of the powerfulness of DL in image classification, there has
been a surge of interest in object detection techniques utilizing DL methodologies in recent years. Pham et al. (2019) conducted a re-
view encompassing various object detection methods in this domain, especially those relying on hand-crafted features and automated
feature extraction by the CNNs. In supervised object detection, ML and DL have been intensively investigated and applied, examples
have been reviewed in (Lone and Pais, 2022). Moreover, CNNs provide higher performance in this task than other conventional algo-
rithms with hand-crafted FE requirements (e.g., KNN, SVM) (Pham et al., 2019). Some object-based lithological classification works
using MSI and ML models revealed good mapping results in both semi-arid with exposed lithology (Serbouti et al., 2021), and vege-
tated terrains (Grebby et al., 2016). Indeed, object based models remain underutilized, despite their potential to yield promising re-
sults in hyperspectral data classification (Booysen et al., 2019). The application of these object-based models is highly recommended
in HLM-MP for both FE and classification tasks.

22
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

7. Summary and conclusions

Over the years, researchers made remarkable progress in using the potential HSI to extract valuable insights about the Earth's sur-
face composition. Furthermore, the ability to classify lithological units accurately aids in understanding geological formations and
structures, which is crucial for successful resource management and planning. As well, accurate mineral identification can lead to
more targeted and efficient exploration activities, reducing costs and minimizing environmental impact. The main conclusions based
on the current review are:
• Spectral-based techniques (SAM, LSU, MTMF, and CEM) have been employed extensively and continuously in HLM-MP studies,
due to their simple implementation. Whereas, Sub-pixel algorithms such as MTMF and CEM show a high ability in mineral
detection compared to other RS techniques, which can be explained by their partial spectral unmixing base.
• Machine learning has demonstrated its capability in enhancing the accuracy and efficiency of mineral prospecting and
lithological mapping. Indeed, the aptitude to detect subtle spectral signatures and discriminate between different mineral
assemblages in complex geological settings provides a valuable advantage in identifying untapped resources. Subsequently, the
DL model's ability can provide great aptitude to map deep features in the HSI data, then an accurate tool to outcropping
mineral phases. However, cloud computing and machine learning-based hyperspectral analysis opens up new possibilities for
uncovering previously undiscovered mineral deposits.
• TIR and SWIR are most sensitive spectral regions to geological variations. There is a need in covering more spectral
information in the TIR region, which is not covered by the majority of the current hyperspectral sensors.
• In terms of data acquisition, high-resolution data acquisition remains a considerable limitation for accurate lithological and
mineral detection and mapping. The High-resolution hyperspectral data acquired using airborne sensors is more suitable for
geological data acquisition, despite their relatively higher cost they can outperform the majority of the new satellite data.
Indeed, the implementation of supervised ML models in HLM-MP is faced with the issue of the labeled data scarcity, which limit
the current advancements in the research field, even though with the availability of global coverage HS data, thus, unsupervised and
semi-supervised ML-based and SU-based algorithms are proposed as alternative algorithms.
The integration of hyperspectral remote sensing data with other geospatial datasets and ground truth information such as geology,
geochemistry, and geophysics allows for a more comprehensive understanding of the earth surface lithology and mineralogy, facili-
tating better-informed decisions in mineral exploration projects. Additionally, the accurate processing of HSI RS and multisource
geospatial integrated data using Knowledge- and Data-driven techniques will have a major influence on the mineral potential map-
ping findings.

CRediT authorship contribution statement

Soufiane Hajaj: Conceptualization, Formal analysis, Methodology, Validation, Writing – original draft, Writing – review & edit-
ing. Abderrazak El Harti: Formal analysis, Methodology, Supervision. Amin Beiranvand Pour: Conceptualization, Formal analy-
sis, Methodology, Writing – original draft. Amine Jellouli: Conceptualization, Methodology, Writing – original draft. Zakaria
Adiri: Formal analysis, Methodology, Writing – original draft. Mazlan Hashim: Formal analysis, Methodology, Writing – original
draft.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Data availability
No data was used for the research described in the article.

Acknowledgement
We extend our sincere appreciation to Prof. Mohamed Khalil Bensalah for his expert guidance in overseeing this publication. The
authors would like to acknowledge the support from the Remote Sensing Department of the National Office of Hydrocarbons and
Mines (ONHYM), Morocco. As well as to the anonymous reviewers for their perceptive remarks and recommendations, that have en-
riched the quality of our work.

References
Abedini, M., Ziaii, M., Timkin, T., Pour, A.B., 2023. Machine learning (ML)-Based copper mineralization prospectivity mapping (MPM) using mining geochemistry
method and remote sensing satellite data. Rem. Sens. 15 (15), 3708.
Aburaed, N., Alkhatib, M.Q., Marshall, S., Zabalza, J., Al Ahmad, H., 2023. A review of spatial enhancement of hyperspectral remote sensing imaging techniques. IEEE
J. Sel. Top. Appl. Earth Obs. Rem. Sens.
Adão, T., Hruška, J., Pádua, L., Bessa, J., Peres, E., Morais, R., Sousa, J.J., 2017. Hyperspectral imaging: a review on UAV-based sensors, data processing and
applications for agriculture and forestry. Rem. Sens. 9 (11), 1110.
Adiri, Z., El Harti, A., Jellouli, A., Maacha, L., Azmi, M., Zouhair, M., Bachaoui, E.M., 2020. Mapping copper mineralization using EO-1 Hyperion data fusion with
Landsat 8 OLI and Sentinel-2A in Moroccan Anti-Atlas. Geocarto Int. 35 (7), 781–800.
Agar, B., Coulter, D., 2007. Remote sensing for mineral exploration–A decade perspective 1997-2007. In: Proceedings Proceedings of Exploration, vol. 7, pp. 109–136.

23
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Agrawal, N., Govil, H., 2023. A deep residual convolutional neural network for mineral classification. Adv. Space Res. 71 (8), 3186–3202.
Agrawal, N., Govil, H., Chatterjee, S., Mishra, G., Mukherjee, S., 2022. Evaluation of machine learning techniques with AVIRIS-NG dataset in the identification and
mapping of minerals. Adv. Space Res.
Agrawal, N., Govil, H., Mishra, G., Gupta, M., Srivastava, P.K., 2023. Evaluating the performance of PRISMA shortwave infrared imaging sensor for mapping
hydrothermally altered and weathered minerals using the machine learning paradigm. Rem. Sens. 15 (12), 3133.
Agterberg, F., Bonham-Carter, G., Cheng, Q.-m., Wright, D., 1993. Weights of evidence modeling and weighted logistic regression for mineral potential mapping.
Comput. Geol. 25, 13–32.
Aravinth, J., Nath, B., Subramanian, M.S., Bulusu, R.V., Monish, P., 2019. Machine learning based detection of zinc mineralization North India using hyperspectral
image processing. In: Proceedings 2019 International Conference on Communication and Electronics Systems (ICCES). IEEE, pp. 1777–1781.
Asadzadeh, S., de Souza Filho, C.R., 2016. A review on spectral processing methods for geological remote sensing. Int. J. Appl. Earth Obs. Geoinf. 47, 69–90.
Aslett, Z., Taranik, J.V., Riley, D.N., 2018. Mapping rock forming minerals at Boundary Canyon, Death Valey National Park, California, using aerial SEBASS thermal
infrared hyperspectral image data. Int. J. Appl. Earth Obs. Geoinf. 64, 326–339.
Awad, M.E., Amer, R., López-Galindo, A., El-Rahmany, M.M., del Moral, L.F.G., Viseras, C., 2018. Hyperspectral remote sensing for mapping and detection of Egyptian
kaolin quality. Appl. Clay Sci. 160, 249–262.
Ayhan, B., Kwan, C., 2017. Application of deep belief network to land cover classification using hyperspectral images. In: Proceedings Advances in Neural Networks-
ISNN 2017: 14th International Symposium, ISNN 2017, Sapporo, Hakodate, and Muroran, Hokkaido, Japan, June 21–26, 2017, Proceedings, Part I 14. Springer, pp.
269–276.
Ayma Quirita, V.A., da Costa, G.A.O.P., Beltrán, C., 2022. A distributed n-findr cloud computing-based solution for endmembers extraction on large-scale hyperspectral
remote sensing data. Rem. Sens. 14 (9), 2153.
Banerjee, B.P., Raval, S., Cullen, P., 2020. UAV-hyperspectral imaging of spectrally complex environments. Int. J. Rem. Sens. 41 (11), 4136–4159.
Bedini, E., 2012. Mapping alteration minerals at Malmbjerg molybdenum deposit, central East Greenland, by Kohonen self-organizing maps and matched filter analysis
of HyMap data. Int. J. Rem. Sens. 33 (4), 939–961.
Bedini, E., 2017. The use of hyperspectral remote sensing for mineral exploration: a review. Journal of Hyperspectral Remote Sensing 7 (4), 189–211.
Bedini, E., Chen, J., 2020. Application of PRISMA satellite hyperspectral imagery to mineral alteration mapping at Cuprite, Nevada, USA. J Journal of Hyperspectral
Remote Sensing v 10 (2), 87–94.
Bedini, E., Chen, J., 2022. Prospection for economic mineralization using PRISMA satellite hyperspectral remote sensing imagery: an example from central East
Greenland. Journal of Hyperspectral Remote Sensing v12 3, 124–130.
Bedini, E., Van Der Meer, F., Van Ruitenbeek, F., 2009. Use of HyMap imaging spectrometer data to map mineralogy in the Rodalquilar caldera, southeast Spain. Int. J.
Rem. Sens. 30 (2), 327–348.
Benhalouche, F.Z., Benabbou, O., Karoui, M.S., Kebir, L.W., Bennia, A., Deville, Y., 2022. Minerals detection and mapping in the Southwestern Algeria Gara-djebilet
region with a multistage informed NMF-based unmixing approach using prisma remote sensing hyperspectral data. In: Proceedings IGARSS 2022-2022 IEEE
International Geoscience and Remote Sensing Symposium. IEEE, pp. 6422–6425.
Bhadra, B., Pathak, S., Nanda, D., Gupta, A., Rao, S.S., 2020. Spectral characteristics of talc and mineral abundance mapping in the Jahazpur Belt of Rajasthan, India
using AVIRIS-NG data. Int. J. Rem. Sens. 41 (22), 8757–8777.
Bioucas-Dias, J.M., Plaza, A., Dobigeon, N., Parente, M., Du, Q., Gader, P., Chanussot, J., 2012. Hyperspectral unmixing overview: geometrical, statistical, and sparse
regression-based approaches. IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens. 5 (2), 354–379.
Blaschke, T., Hay, G.J., Kelly, M., Lang, S., Hofmann, P., Addink, E., Feitosa, R.Q., Van der Meer, F., Van der Werff, H., Van Coillie, F., 2014. Geographic object-based
image analysis–towards a new paradigm. ISPRS J. Photogrammetry Remote Sens. 87, 180–191.
Blum, A., Mitchell, T., 1998. Combining labeled and unlabeled data with co-training. In: Proceedings Proceedings of the Eleventh Annual Conference on Computational
Learning Theory. pp. 92–100.
Boardman, J., Kruse, F., 1994. Automated spectral analysis: a geologic example using AVIRIS data. In: Proceedings Proceedings ERIM Tenth Thematic Conference on
Geologic Remote Sensing. Environmental Research Institute of Michigan, Ann Arbor, Michigan, pp. 407–418.
Bonham-Carter, G., 1994. Geographic Information Systems for Geoscientists: Modelling with GIS, vol. 13. Elsevier.
Booysen, R., Gloaguen, R., Lorenz, S., Zimmermann, R., Andreani, L., Nex, P.A., 2019. The potential of multi-sensor remote sensing mineral exploration: examples from
Southern Africa. In: Proceedings IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 6027–6030.
Booysen, R., Jackisch, R., Lorenz, S., Zimmermann, R., Kirsch, M., Nex, P.A., Gloaguen, R., 2020. Detection of REEs with lightweight UAV-based hyperspectral imaging.
Sci. Rep. 10 (1), 17450.
Booysen, R., Lorenz, S., Thiele, S.T., Fuchsloch, W.C., Marais, T., Nex, P.A., Gloaguen, R., 2022. Accurate hyperspectral imaging of mineralised outcrops: an example
from lithium-bearing pegmatites at Uis, Namibia. Rem. Sens. Environ. 269, 112790.
Breiman, L., 2001. Random Forests: Machine Learning, vol. 45. pp. 5–32.
Bruzzone, L., Chi, M., Marconcini, M., 2006. A novel transductive SVM for semisupervised classification of remote-sensing images. IEEE Trans. Geosci. Rem. Sens. 44
(11), 3363–3373.
Buckley, S.J., Kurz, T.H., Howell, J.A., Schneider, D., 2013. Terrestrial lidar and hyperspectral data fusion products for geological outcrop analysis. Comput. Geosci. 54,
249–258.
Cai, J., Wang, L., Huang, Y., Chen, D., Liu, H., 2020. Formation lithology identification technology along railway in complex and dangerous mountainous area based on
hyperspectral. In: Proceedings IOP Conference Series: Materials Science and Engineering, vol. 780. IOP Publishing, 042049.
Camps-Valls, G., Marsheva, T.V.B., Zhou, D., 2007. Semi-supervised graph-based hyperspectral image classification. IEEE Trans. Geosci. Rem. Sens. 45 (10),
3044–3054.
Canbaz, O., Gürsoy, Ö., Karaman, M., Çalışkan, A.B., Gökce, A., 2021. Hydrothermal alteration mapping using EO-1 Hyperion hyperspectral data in Kösedağ, Central-
Eastern Anatolia (Sivas-Turkey). Arabian J. Geosci. 14 (21), 2245.
Cardoso-Fernandes, J., Santos, D., Lima, A., Teodoro, A.C., 2022. Assessing the PRISMA potential for mineral exploration to vector low-grade lithium deposits. In:
Proceedings IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 5985–5988.
Carranza, E.J.M., Hale, M., 2003. Evidential belief functions for data-driven geologically constrained mapping of gold potential, Baguio district, Philippines. Ore Geol.
Rev. 22 (1–2), 117–132.
Carrino, T.A., Crósta, A.P., Toledo, C.L.B., Silva, A.M., 2018. Hyperspectral remote sensing applied to mineral exploration in southern Peru: a multiple data integration
approach in the Chapi Chiara gold prospect. Int. J. Appl. Earth Obs. Geoinf. 64, 287–300.
Caruso, A., Clarke, K., Tiddy, C., Lewis, M., 2021. Airborne hyperspectral characterisation of hydrothermal alteration in a regolith-dominated terrain, southern Gawler
Ranges, South Australia. Aust. J. Earth Sci. 68 (4), 590–608.
Chabrillat, S., Brell, M., Segl, K., Foerster, S., Guanter, L., Schickling, A., Storch, T., Honold, H.-P., Fischer, S., 2021. The EnMAP satellite-mission status and science
preparatory activities. In: Proceedings 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. IEEE, pp. 123–125.
Chabrillat, S., Milewski, R., Ward, K., Foerster, S., Guillaso, S., Loy, C., Ben-Dor, E., Tziolas, N., Schmid, T., van Wesemael, B., 2023. Monitoring soil properties using
EnMAP spaceborne imaging spectroscopy mission. In: Proceedings IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp.
1130–1133.
Chakraborty, R., Kereszturi, G., Pullanagari, R., Durance, P., Ashraf, S., Anderson, C., 2022. Mineral prospecting from biogeochemical and geological information using
hyperspectral remote sensing-Feasibility and challenges. J. Geochem. Explor. 232, 106900.
Chang, C.-I., 2003. Hyperspectral Imaging: Techniques for Spectral Detection and Classification. Springer Science & Business Media.
Chang, C.-I., Wu, C.-C., Liu, K.-H., Chen, H.-M., Chen, C.C.-C., Wen, C.-H., 2014. Progressive band processing of linear spectral unmixing for hyperspectral imagery.
IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens. 8 (6), 2583–2597.
Chapman, J.W., Thompson, D.R., Helmlinger, M.C., Bue, B.D., Green, R.O., Eastwood, M.L., Geier, S., Olson-Duvall, W., Lundeen, S.R., 2019. Spectral and radiometric
calibration of the next generation airborne visible infrared spectrometer (AVIRIS-NG). Rem. Sens. 11 (18), 2129.

24
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Chattoraj, S.L., Prasad, G., Sharma, R.U., van der Meer, F.D., Guha, A., Pour, A.B., 2020. Integration of remote sensing, gravity and geochemical data for exploration of
Cu-mineralization in Alwar basin, Rajasthan, India. Int. J. Appl. Earth Obs. Geoinf. 91, 102162.
Chen, W., Li, X., Qin, X., Wang, L., 2024. Hyperspectral remote sensing inversion of mineral abundance based on sparse unmixing method. Remote Sensing Intelligent
Interpretation for Geology: From Perspective of Geological Exploration 211–229. . Springer.
Chirico, R., Mondillo, N., Laukamp, C., Mormone, A., Di Martire, D., Novellino, A., Balassone, G., 2022. Mapping hydrothermal and supergene alteration zones
associated with carbonate-hosted Zn-Pb deposits by using PRISMA satellite imagery supported by field-based hyperspectral data, mineralogical and geochemical
analysis. Ore Geol. Rev. 105244.
Clark, R.N., Rencz, A.N., 1999. Spectroscopy of rocks and minerals, and principles of spectroscopy: manual of remote sensing. 3 (11), 3–58.
Cloutis, E.A., 1996. Review article hyperspectral geological remote sensing: evaluation of analytical techniques. Int. J. Rem. Sens. 17 (12), 2215–2242.
Combe, J.-P., Le Mouélic, S., Sotin, C., Gendrin, A., Mustard, J., Le Deit, L., Launeau, P., Bibring, J.-P., Gondet, B., Langevin, Y., 2008. Analysis of OMEGA/Mars express
data hyperspectral data using a multiple-endmember linear spectral unmixing model (MELSUM): methodology and first results. Planet. Space Sci. 56 (7), 951–975.
Contreras, C., Khodadadzadeh, M., Ghamisi, P., Gloaguen, R., 2019. Mineral mapping of drill core hyperspectral data with extreme learning machines. In: Proceedings
IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 2686–2689.
Crosta, A. P., and Moore, J., Rehancement of LANDSAT Thematic Mapper imagery for residual soil mapping in SW Minas Gerais State, Brazil: a prospecting case history
in Greenstone Belt terrain, in Proceedings Thematic conference on remote sensing for exploration geology. Methods, integration, solutions. 71990, p. 1173-1187.
Crowley, J.K., Brickey, D.W., Rowan, L.C., 1989. Airborne imaging spectrometer data of the Ruby Mountains, Montana: mineral discrimination using relative
absorption band-depth images. Rem. Sens. Environ. 29 (2), 121–134.
Dabiri, Z., Lang, S., 2018. Comparison of independent component analysis, principal component analysis, and minimum noise fraction transformation for tree species
classification using APEX hyperspectral imagery. ISPRS Int. J. Geo-Inf. 7 (12), 488.
De Boissieu, F., Sevin, B., Cudahy, T., Mangeas, M., Chevrel, S., Ong, C., Rodger, A., Maurizot, P., Laukamp, C., Lau, I., 2018. Regolith-geology mapping with support
vector machine: a case study over weathered Ni-bearing peridotites, New Caledonia. Int. J. Appl. Earth Obs. Geoinf. 64, 377–385.
De La Rosa, R., Khodadadzadeh, M., Tusa, L., Kirsch, M., Gisbert, G., Tornos, F., Tolosana-Delgado, R., Gloaguen, R., 2021. Mineral quantification at deposit scale using
drill-core hyperspectral data: a case study in the Iberian Pyrite Belt. Ore Geol. Rev. 139, 104514.
Deng, K., Zhao, H., Li, N., Wei, W., 2021. Identification of minerals in hyperspectral imagery based on the attenuation spectral absorption index vector using a
multilayer perceptron. Remote Sensing Letters 12 (5), 449–458.
Dong, X., Gan, F., Li, N., Zhang, S., Li, T., 2022. Mineral mapping in the Duolong porphyry and epithermal ore district, Tibet, using the Gaofen-5 satellite hyperspectral
remote sensing data. Ore Geol. Rev. 105222.
Dópido, I., Li, J., Marpu, P.R., Plaza, A., Dias, J.M.B., Benediktsson, J.A., 2013. Semisupervised self-learning for hyperspectral image classification. IEEE Trans. Geosci.
Rem. Sens. 51 (7), 4032–4044.
dos Reis Salles, R., de Souza Filho, C.R., Cudahy, T., Vicente, L.E., Monteiro, L.V.S., 2017. Hyperspectral remote sensing applied to uranium exploration: a case study at
the Mary Kathleen metamorphic-hydrothermal U-REE deposit, NW, Queensland, Australia. J. Geochem. Explor. 179, 36–50.
Elith, J., Phillips, S.J., Hastie, T., Dudík, M., Chee, Y.E., Yates, C.J., 2011. A statistical explanation of MaxEnt for ecologists: diversity and distributions. 17 (1), 43–57.
Esmaeilia, M., Fathianpour, N., Soltani-Mohammadi, S., 2024. PRISMA hyperspectral imagery for mapping alteration zones associated with Kuhpanj porphyry copper
deposit, Southern Iran. European Journal of Remote Sensing.
Fan, Y., Wan, Y., Wang, H., Yang, X., Liang, M., Pan, C., Zhang, S., Wang, W., Tan, F., 2021. Application of an airborne hyper-spectral survey system CASI/SASI in the
gold-silver-lead-zinc ore district of Huaniushan. Gansu, China: Geol. Croat. 74 (1), 73–83.
Fang, Y., Xiao, Y., Liang, S., Ji, Y., Chen, H., 2022. Lithological classification by PCA-QPSO-LSSVM method with thermal infrared hyper-spectral data. J. Appl. Remote
Sens. 16 (4), 44515. 044515.
Fassnacht, F.E., Neumann, C., Förster, M., Buddenbaum, H., Ghosh, A., Clasen, A., Joshi, P.K., Koch, B., 2014. Comparison of feature reduction algorithms for classifying
tree species with hyperspectral data on three central European test sites. IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens. 7 (6), 2547–2561.
Feng, J., Rogge, D., Rivard, B., 2018. Comparison of lithological mapping results from airborne hyperspectral VNIR-SWIR, LWIR and combined data. Int. J. Appl. Earth
Obs. Geoinf. 64, 340–353.
Feng, X.-R., Li, H.-C., Wang, R., Du, Q., Jia, X., Plaza, A., 2022. Hyperspectral unmixing based on nonnegative matrix factorization: a comprehensive review. IEEE J. Sel.
Top. Appl. Earth Obs. Rem. Sens. 15, 4414–4436.
Ferrier, G., Ganas, A., Pope, R., Jo Miles, A., 2019. Prospectivity mapping for epithermal deposits of Western Milos using a fuzzy multi criteria evaluation approach
parameterized by airborne hyperspectral remote sensing data. Geosciences 9 (3), 116.
Fu, H., Cheng, Q., Jing, L., Ge, Y., 2021. Deep learning-based hydrothermal alteration mapping using GaoFen-5 hyperspectral data in the Duolong Ore District, Western
Tibet, China. J. Appl. Remote Sens. 15 (4), 44512. 044512.
Fu, Y., Cheng, Q., Jing, L., Ye, B., Fu, H., 2023. Mineral prospectivity mapping of porphyry copper deposits based on remote sensing imagery and geochemical data in
the Duolong ore district, Tibet. Rem. Sens. 15 (2), 439.
Fung, C.C., Wong, K.W., 2005. Comparing the performance of different neural networks architectures for the prediction of mineral prospectivity. In: Proceedings 2005
International Conference on Machine Learning and Cybernetics, vol. 1. IEEE, pp. 394–398.
Gaffey, S.J., 1986. Spectral reflectance of carbonate minerals in the visible and near infrared (0.35-2.55 microns); calcite, aragonite, and dolomite. Am. Mineral. 71
(1–2), 151–162.
George, E.B., Ternikar, C.R., Ghosh, R., Kumar, D.N., Gomez, C., Ahmad, T., Sahadevan, A.S., Gupta, P.K., Misra, A., 2022. Assessment of spectral reduction techniques
for endmember extraction in unmixing of hyperspectral images. Adv. Space Res.
Ghamisi, P., Rasti, B., Yokoya, N., Wang, Q., Hofle, B., Bruzzone, L., Bovolo, F., Chi, M., Anders, K., Gloaguen, R., 2019. Multisource and multitemporal data fusion in
remote sensing: a comprehensive review of the state of the art. IEEE Geoscience and Remote Sensing Magazine 7 (1), 6–39.
Gladwell, D., Lett, R., Lawrence, P., 1983. Application of reflectance spectrometry to mineral exploration using portable radiometers. Econ. Geol. 78 (4), 699–710.
Govil, H., Mishra, G., Gill, N., Taloor, A., Diwan, P., 2021. Mapping hydrothermally altered minerals and gossans using hyperspectral data in eastern kumaon himalaya,
India. Applied Computing and Geosciences 9, 100054.
Grebby, S., Field, E., Tansey, K., 2016. Evaluating the use of an object-based approach to lithological mapping in vegetated terrain. Rem. Sens. 8 (10), 843.
Green, A.A., Berman, M., Switzer, P., Craig, M.D., 1988. A transformation for ordering multispectral data in terms of image quality with implications for noise removal.
IEEE Trans. Geosci. Rem. Sens. 26 (1), 65–74.
Guha, A., 2020. Mineral exploration using hyperspectral data. Hyperspectral Remote Sensing 293–318. . Elsevier.
Guha, A., Chatterjee, S., Oommen, T., Kumar, K.V., Roy, S.K., 2021a. Synergistic use of ASTER, L-band ALOS PALSAR, and hyperspectral AVIRIS-NG data for
exploration of lode type gold deposit–a study in Hutti Maski Schist Belt, India. Ore Geol. Rev. 128, 103818.
Guha, A., Kumar Ghosh, U., Sinha, J., Pour, A.B., Bhaisal, R., Chatterjee, S., Kumar Baranval, N., Rani, N., Kumar, K.V., Rao, P.V., 2021b. Potentials of airborne
hyperspectral AVIRIS-NG data in the exploration of base metal deposit—a study in the parts of Bhilwara, Rajasthan. Rem. Sens. 13 (11), 2101.
Guha, A., Mondal, S., Chatterjee, S., Kumar, K.V., 2022. Airborne imaging spectroscopy of igneous layered complex and their mapping using different spectral
enhancement conjugated support vector machine models. Geocarto Int. 37 (2), 349–365.
Guo, X., Li, P., Li, J., 2021. Lithological mapping using EO-1 Hyperion hyperspectral data and semisupervised self-learning method. J. Appl. Remote Sens. 15 (3), 32209.
032209.
Hajaj, S., El Harti, A., Jellouli, A., Pour, A.B., Himyari, S.M., Hamzaoui, A., Bensalah, M.K., Benaouiss, N., Hashim, M., 2023a. HyMap imagery for copper and
manganese prospecting in the east of Ameln valley shear zone (Kerdous inlier, western Anti-Atlas, Morocco). Spatial Sci. 1–22.
Hajaj, S., El Harti, A., Jellouli, A., Pour, A.B., Mnissar Himyari, S., Hamzaoui, A., Hashim, M., 2023b. Evaluating the performance of machine learning and deep learning
techniques to HyMap imagery for lithological mapping in a semi-arid region: case study from western Anti-Atlas, Morocco. Minerals 13 (6), 766.
HajiRassouliha, A., Taberner, A.J., Nash, M.P., Nielsen, P.M., 2018. Suitability of recent hardware accelerators (DSPs, FPGAs, and GPUs) for computer vision and image
processing algorithms. Signal Process. Image Commun. 68, 101–119.
Harris, J., Ponomarev, P., Shang, J., Rogge, D., 2006. Noise reduction and best band selection techniques for improving classification results using hyperspectral data:

25
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

application to lithological mapping in Canada’s Arctic. Can. J. Rem. Sens. 32 (5), 341–354.
Huang, G.-B., Zhu, Q.-Y., Siew, C.-K., 2006. Extreme learning machine: theory and applications. Neurocomputing 70 (1–3), 489–501.
Huang, Z., Zheng, J., 2018. Alteration minerals extraction using airborne hyperspectral data Casi and Sasi in Wuyi metallogenic belt, China: the international archives
of the photogrammetry. Remote Sensing and Spatial Information Sciences 42, 601–605.
Hunt, G.R., 1977. Spectral signatures of particulate minerals in the visible and near infrared. Geophysics 42 (3), 501–513.
Hunt, G. R., 2017, Spectroscopic properties of rocks and minerals: handbook of physical properties of rocks, v. 1, p. 295-385.
Jackisch, R., Madriz, Y., Zimmermann, R., Pirttijärvi, M., Saartenoja, A., Heincke, B.H., Salmirinne, H., Kujasalo, J.-P., Andreani, L., Gloaguen, R., 2019. Drone-borne
hyperspectral and magnetic data integration: Otanmäki Fe-Ti-V deposit in Finland. Rem. Sens. 11 (18), 2084.
Jain, R., Bhu, H., Kumar, H., Purohit, R., 2022. Integration of multi-sensor remote sensing, geological and geochemical data for delineation of Pb–Zn bearing carbonates
of Middle Aravalli group in Zawar–Dungarpur Belt, NW India. Geocarto Int. 37 (27), 17165–17199.
Jain, R., Sharma, R.U., 2019. Airborne hyperspectral data for mineral mapping in Southeastern Rajasthan, India. Int. J. Appl. Earth Obs. Geoinf. 81, 137–145.
Jin, G., Raich, R., Miller, D.J., 2013. A generative semi-supervised model for multi-view learning when some views are label-free. In: Proceedings 2013 IEEE
International Conference on Acoustics, Speech and Signal Processing. IEEE, pp. 3302–3306.
Kale, K.V., Solankar, M.M., Nalawade, D.B., Dhumal, R.K., Gite, H.R., 2017. A research review on hyperspectral data processing and analysis algorithms. Proc. Natl.
Acad. Sci., India, Sect. A 87, 541–555.
Kayet, N., Pathak, K., Chakrabarty, A., Sahoo, S., 2018. Mapping the distribution of iron ore minerals and spatial correlation with environmental variables in hilltop
mining areas. Environ. Earth Sci. 77, 1–14.
Kereszturi, G., Pullanagari, R.R., Mead, S., Schaefer, L.N., Procter, J., Schleiffarth, W.K., Kennedy, B., 2018. Geological mapping of hydrothermal alteration on
volcanoes from multi-sensor platforms. In: Proceedings IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 220–223.
Khosravi, V., Shirazi, A., Shirazy, A., Hezarkhani, A., Pour, A.B., 2021. Hybrid fuzzy-analytic hierarchy process (AHP) model for porphyry copper prospecting in
simorgh area, eastern lut block of Iran: mining. 2 (1), 1–12.
Koirala, B., Khodadadzadeh, M., Contreras, C., Zahiri, Z., Gloaguen, R., Scheunders, P., 2019. A supervised method for nonlinear hyperspectral unmixing. Rem. Sens. 11
(20), 2458.
Krupnik, D., Khan, S., 2019. Close-range, ground-based hyperspectral imaging for mining applications at various scales: review and case studies. Earth Sci. Rev. 198,
102952.
Kumar, C., Chatterjee, S., Oommen, T., 2020a. Mapping hydrothermal alteration minerals using high-resolution AVIRIS-NG hyperspectral data in the Hutti-Maski gold
deposit area, India. Int. J. Rem. Sens. 41 (2), 794–812.
Kumar, C., Chatterjee, S., Oommen, T., Guha, A., 2020b. Automated lithological mapping by integrating spectral enhancement techniques and machine learning
algorithms using AVIRIS-NG hyperspectral data in Gold-bearing granite-greenstone rocks in Hutti, India. Int. J. Appl. Earth Obs. Geoinf. 86, 102006.
Kumar, H., Sharma, A., Rajawat, A., 2018. Applications of imaging spectroscopy for non-metallic mineral exploration: the international archives of the
photogrammetry. Remote Sensing and Spatial Information Sciences 42, 835–838.
Kuras, A., Heincke, B.H., Salehi, S., Mielke, C., Köllner, N., Rogass, C., Altenberger, U., Burud, I., 2022. Integration of hyperspectral and magnetic data for geological
characterization of the niaqornarssuit ultramafic complex in west-Greenland. Rem. Sens. 14 (19), 4877.
Kurz, T.H., Buckley, S.J., Howell, J.A., Schneider, D., 2009. Close range hyperspectral and LiDAR data integration for geological outcrop analysis. In: Proceedings 2009
First Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing. IEEE, pp. 1–4.
Kurz, T.H., San Miguel, G., Dubucq, D., Kenter, J., Miegebielle, V., Buckley, S.J., 2022. Quantitative mapping of dolomitization using close-range hyperspectral imaging:
Kimmeridgian carbonate ramp, Alacón, NE Spain. Geosphere 18 (2), 780–799.
Laukamp, C., 2022. Geological mapping using mineral absorption feature-Guided band-ratios applied to prisma satellite hyperspectral level 2D imagery. In: Proceedings
IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 5981–5984.
Laukamp, C., Cudahy, T., Thomas, M., Jones, M., Cleverley, J., Oliver, N., 2011. Hydrothermal mineral alteration patterns in the Mount Isa Inlier revealed by airborne
hyperspectral data. Aust. J. Earth Sci. 58 (8), 917–936.
Li, Q., Chen, G., Luo, L., 2023. Mineral prospectivity mapping using attention–based convolutional neural network. Ore Geol. Rev. 105381.
Libeesh, N., Arivazhagan, S., 2023. Spectral pathways for effective delineation of mafic-ultramafic rocks by using PRISMA hyperspectral data. Remote Sens. Appl.:
Society and Environment 30, 100955.
Lin, N., Chen, Y., Liu, H., Liu, H., 2021. A comparative study of machine learning models with hyperparameter optimization algorithm for mapping mineral
prospectivity. Minerals 11 (2), 159.
Lin, N., Fu, J., Jiang, R., Li, G., Yang, Q., 2023. Lithological classification by hyperspectral images based on a two-layer XGBoost model, combined with a greedy
algorithm. Rem. Sens. 15 (15), 3764.
Lin, N., Liu, H., Li, G., Wu, M., Li, D., Jiang, R., Yang, X., 2022. Extraction of mineralized indicator minerals using ensemble learning model optimized by SSA based on
hyperspectral image. Open Geosci. 14 (1), 1444–1465.
Liu, H., Wu, K., Xu, H., Xu, Y., 2021. Lithology classification using TASI thermal infrared hyperspectral data with convolutional neural networks. Rem. Sens. 13 (16),
3117.
Liu, P., 2020. Application of cloud computing in geological exploration. In: Proceedings IOP Conference Series: Materials Science and Engineering, vol. 750. IOP
Publishing, 012159.
Lodhi, V., Chakravarty, D., Mitra, P., 2018. Hyperspectral imaging for earth observation: platforms and instruments. J. Indian Inst. Sci. 98 (4), 429–443.
Loizzo, R., Ananasso, C., Guarini, R., Lopinto, E., Candela, L., Pisani, A., 2016. The prisma hyperspectral mission. In: Proceedings Proceedings of the Living Planet
Symposium. pp. 9–13. Prague, Czech Republic.
Lone, Z.A., Pais, A.R., 2022. Object Detection in Hyperspectral Images: Digital Signal Processing. 103752.
Lorenz, S., Ghamisi, P., Kirsch, M., Jackisch, R., Rasti, B., Gloaguen, R., 2021. Feature extraction for hyperspectral mineral domain mapping: a test of conventional and
innovative methods. Rem. Sens. Environ. 252, 112129.
Lorenz, S., Kirsch, M., Zimmermann, R., Tusa, L., Möckel, R., Chamberland, M., Gloaguen, R., 2018a. Long-wave hyperspectral imaging for lithological mapping: a case
study. In: Proceedings IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 1620–1623.
Lorenz, S., Salehi, S., Kirsch, M., Zimmermann, R., Unger, G., Vest Sørensen, E., Gloaguen, R., 2018b. Radiometric correction and 3D integration of long-range ground-
based hyperspectral imagery for mineral exploration of vertical outcrops. Rem. Sens. 10 (2), 176.
Lu, B., Dao, P.D., Liu, J., He, Y., Shang, J., 2020. Recent advances of hyperspectral imaging technology and applications in agriculture. Rem. Sens. 12 (16), 2659.
Lu, D., Weng, Q., 2007. A survey of image classification methods and techniques for improving classification performance. Int. J. Rem. Sens. 28 (5), 823–870.
Lypaczewski, P., Rivard, B., Gaillard, N., Perrouty, S., Piette-Lauzière, N., Bérubé, C.L., Linnen, R.L., 2019. Using hyperspectral imaging to vector towards
mineralization at the Canadian Malartic gold deposit, Québec, Canada. Ore Geol. Rev. 111, 102945.
Malehmir, A., Dynesius, L., Paulusson, K., Paulusson, A., Johansson, H., Bastani, M., Wedmark, M., Marsden, P., 2017. The potential of rotary-wing UAV-based
magnetic surveys for mineral exploration: a case study from central Sweden. Lead. Edge 36 (7), 552–557.
Manolakis, D., Siracusa, C., Shaw, G., 2001. Hyperspectral subpixel target detection using the linear mixing model. IEEE Trans. Geosci. Rem. Sens. 39 (7), 1392–1409.
McCormick, C.A., Corlett, H., Stacey, J., Hollis, C., Feng, J., Rivard, B., Omma, J.E., 2021. Shortwave infrared hyperspectral imaging as a novel method to elucidate
multi-phase dolomitization, recrystallization, and cementation in carbonate sedimentary rocks. Sci. Rep. 11 (1), 21732.
Meer, F.v. d., Jong, S.D., Bakker, W., 2001. Imaging Spectrometry: Basic Analytical Techniques: Imaging Spectrometry: Basic Principles and Prospective Applications.
pp. 17–61.
Meyer, J.M., Holley, E., Kokaly, R.F., Swayze, G.A., Hoefen, T.M., 2021. Mineral mapping of the battle mountain district, Nevada, USA, using AVIRIS-classic and
SpecTIR Inc. AisaFENIX 1K imaging spectrometer datasets. In: Proceedings 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. IEEE, pp.
1859–1862.
Mishra, G., Govil, H., Guha, A., Kumar, H., Kumar, S., Mukherjee, S., 2022a. Comparative evaluation of airborne AVIRIS-NG and spaceborne PRISMA hyperspectral data
in identification and mapping of altered/weathered minerals in Jahazpur, Rajasthan. Adv. Space Res.

26
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Mishra, G., Govil, H., Guha, A., Rajendran, S., 2022b. Identification of key altered/weathered minerals near to the base metal mineral in Jahazpur, India using AVIRIS-
NG data. Adv. Space Res.
Mishra, G., Govil, H., Srivastava, P.K., 2021. Identification of malachite and alteration minerals using airborne AVIRIS-NG hyperspectral data. Quaternary Science
Advances 4, 100036.
Molan, Y.E., Refahi, D., Tarashti, A.H., 2014. Mineral mapping in the Maherabad area, eastern Iran, using the HyMap remote sensing data. Int. J. Appl. Earth Obs.
Geoinf. 27, 117–127.
Mondal, S., Guha, A., Pal, S.K., 2022. Support vector machine-based integration of AVIRIS NG hyperspectral and ground geophysical data for identifying potential zones
for chromite exploration–A study in Tamil Nadu, India. Adv. Space Res.
Mou, L., Ghamisi, P., Zhu, X.X., 2017. Deep recurrent neural networks for hyperspectral image classification. IEEE Trans. Geosci. Rem. Sens. 55 (7), 3639–3655.
Ni, L., Xu, H., Zhou, X., 2020. Mineral identification and mapping by synthesis of hyperspectral VNIR/SWIR and multispectral TIR remotely sensed data with different
classifiers. IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens. 13, 3155–3163.
Pal, M., Rasmussen, T., Porwal, A., 2020. Optimized lithological mapping from multispectral and hyperspectral remote sensing images using fused multi-classifiers.
Rem. Sens. 12 (1), 177.
Paoletti, M., Haut, J., Plaza, J., Plaza, A., 2019. Deep learning classifiers for hyperspectral imaging: a review. ISPRS J. Photogrammetry Remote Sens. 158, 279–317.
Paul, S., Kumar, D.N., 2019. Partial informational correlation-based band selection for hyperspectral image classification. J. Appl. Remote Sens. 13 (4), 46505. 046505.
Peyghambari, S., Zhang, Y., 2021. Hyperspectral remote sensing in lithological mapping, mineral exploration, and environmental geology: an updated review. J. Appl.
Remote Sens. 15 (3), 31501. 031501.
Pham, T.T., Takalkar, M.A., Xu, M., Hoang, D.T., Truong, H., Dutkiewicz, E., Perry, S., 2019. Airborne object detection using hyperspectral imaging: deep learning
review. In: Proceedings Computational Science and its Applications–ICCSA 2019: 19th International Conference, Saint Petersburg, Russia, July 1–4, 2019,
Proceedings, Part I 19. Springer, pp. 306–321.
Pour, A.B., Hashim, M., Hong, J.K., Park, Y., 2019. Lithological and alteration mineral mapping in poorly exposed lithologies using Landsat-8 and ASTER satellite data:
North-eastern Graham Land, Antarctic Peninsula. Ore Geol. Rev. 108, 112–133.
Pour, A.B., Ranjbar, H., Sekandari, M., Abd El-Wahed, M., Hossain, M.S., Hashim, M., Yousefi, M., Zoheir, B., Wambo, J.D.T., Muslim, A.M., 2023. Remote sensing for
mineral exploration. Geospatial Analysis Applied to Mineral Exploration 17–149. . Elsevier.
Pour, A.B., Zoheir, B., Pradhan, B., Hashim, M., 2021. Editorial for the Special Issue: Multispectral and Hyperspectral Remote Sensing Data for Mineral Exploration and
Environmental Monitoring of Mined Areas, vol. 13. MDPI, p. 519.
Qasim, M., Khan, S.D., Haider, R., Rasheed, M.u., 2022. Integration of multispectral and hyperspectral remote sensing data for lithological mapping in Zhob Ophiolite,
Western Pakistan. Arabian J. Geosci. 15 (7), 599.
Qin, J., Zhan, Y., Dai, C., Yao, W., Wu, K., Liu, W., Cao, Y., Zhang, X., Medjadba, Y., Wang, Y., 2019. Geologic body classification of hyperspectral data based on dilated
convolution neural network at Tianshan area. In: Proceedings IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp.
2985–2988.
Qin, Y., Zhang, X., Zhao, Z., Li, Z., Yang, C., Huang, Q., 2021. Coupling relationship analysis of gold content using Gaofen-5 (GF-5) satellite hyperspectral remote
sensing data: a potential method in Chahuazhai gold mining area, Qiubei county, SW China. Rem. Sens. 14 (1), 109.
Rani, K., Guha, A., Kumar, K.V., Bhattacharya, B.K., Pradeep, B., 2020. Potential use of airborne hyperspectral AVIRIS-NG data for mapping proterozoic metasediments
in Banswara, India. J. Geol. Soc. India 95 (2), 152–158.
Rast, M., Painter, T.H., 2019. Earth observation imaging spectroscopy for terrestrial systems: an overview of its history, techniques, and applications of its missions.
Surv. Geophys. 40 (3), 303–331.
Ren, Z., Sun, L., Zhai, Q., 2020. Improved k-means and spectral matching for hyperspectral mineral mapping. Int. J. Appl. Earth Obs. Geoinf. 91, 102154.
Resmini, R., Kappus, M., Aldrich, W., Harsanyi, J., Anderson, M., 1997. Mineral mapping with hyperspectral digital imagery collection experiment (HYDICE) sensor
data at Cuprite, Nevada, USA. Int. J. Rem. Sens. 18 (7), 1553–1570.
Rialland, R., Carrère, V., Marion, R., Soussen, C., 2021. Identification of minerals from hyperspectral imaging based on a fuzzy logic approach. In: Proceedings
Algorithms, Technologies, and Applications for Multispectral and Hyperspectral Imaging XXVII, vol. 11727. SPIE, pp. 121–132.
Salehi, S., Lorenz, S., Vest Sørensen, E., Zimmermann, R., Fensholt, R., Henning Heincke, B., Kirsch, M., Gloaguen, R., 2018. Integration of vessel-based hyperspectral
scanning and 3D-photogrammetry for mobile mapping of steep coastal cliffs in the arctic. Rem. Sens. 10 (2), 175.
Salehi, S., Mielke, C., Rogass, C., 2020. Mapping ultramafic complexes using airborne imaging spectroscopy and spaceborne data in Arctic regions with abundant lichen
cover, a case study from the Niaqornarssuit complex in South West Greenland. European Journal of Remote Sensing 53 (1), 156–175.
Samani, P., Prizomwala, S., Rajawat, A., 2021. Assessing the mineral alteration in Ambaji–Deri Region (Northwestern India) using hyperspectral remote sensing.
Journal of the Indian Society of Remote Sensing 49, 249–257.
Sankey, T.T., McVay, J., Swetnam, T.L., McClaran, M.P., Heilman, P., Nichols, M., 2018. UAV hyperspectral and lidar data and their fusion for arid and semi-arid land
vegetation monitoring. Remote Sensing in Ecology and Conservation 4 (1), 20–33.
Sawant, S.S., Manoharan, P., Loganathan, A., 2021. Band selection strategies for hyperspectral image classification based on machine learning and artificial intelligent
techniques–Survey. Arabian J. Geosci. 14, 1–10.
Sawant, S.S., Prabukumar, M., 2017. Semi-supervised techniques based hyper-spectral image classification: a survey: 2017 Innovations. In: Power and Advanced
Computing Technologies (I-PACT). pp. 1–8.
Schmidt, F., Legendre, M., Le Mouëlic, S., 2014. Minerals detection for hyperspectral images using adapted linear unmixing: LinMin. Icarus 237, 61–74.
Schodlok, M.C., Frei, M., 2020. LWIR hyperspectral mapping of the Gamsberg deposit, Aggeneys, South Africa. In: Proceedings IGARSS 2020-2020 IEEE International
Geoscience and Remote Sensing Symposium. IEEE, pp. 5135–5138.
Sengupta, A., Adhikari, M.D., Maiti, S., Maiti, S.K., Mahanta, P., Bhaumick, S., 2019. Identification and mapping of high-potential iron ore alteration zone across Joda,
Odisha using ASTER and EO-1 hyperion data. Spatial Sci. 64 (3), 491–514.
Serbouti, I., Raji, M., Hakdaoui, M., Pradhan, B., Lee, C.-W., Alamri, A.M., 2021. Pixel and object-based machine learning classification schemes for lithological
mapping enhancement of semi-arid regions using sentinel-2A imagery: a case study of the southern Moroccan meseta. IEEE Access 9, 119262–119278.
Shahmoradi, J., Talebi, E., Roghanchi, P., Hassanalian, M., 2020. A comprehensive review of applications of drone technology in the mining industry. Drones 4 (3), 34.
Shahtakhtinskiy, A., Khan, S., 2020. Quantitative analysis of facies variation using ground-based lidar and hyperspectral imaging in Mississippian limestone outcrop
near Jane, Missouri. Interpretation 8 (2), T365–T378.
Shaik, I., Begum, S., Nagamani, P., Kayet, N., 2021. Characterization and mapping of hematite ore mineral classes using hyperspectral remote sensing technique: a case
study from Bailadila iron ore mining region. SN Appl. Sci. 3, 1–13.
Shebl, A., Abriha, D., Fahil, A.S., El-Dokouny, H.A., Elrasheed, A.A., Csámer, Á., 2023. PRISMA hyperspectral data for lithological mapping in the Egyptian eastern
desert: evaluating the support vector machine, random forest, and XG boost machine learning algorithms. Ore Geol. Rev. 105652.
Shevyrev, S., Carranza, E.J.M., 2022. Application of maximum entropy for mineral prospectivity mapping in heavily vegetated areas of Greater Kurile Chain with
Landsat 8 data. Ore Geol. Rev. 142, 104758.
Shimoni, M., Haelterman, R., Perneel, C., 2019. Hypersectral imaging for military and security applications: combining myriad processing and sensing techniques. IEEE
Geoscience and Remote Sensing Magazine 7 (2), 101–117.
Shirazi, A., Hezarkhani, A., Beiranvand Pour, A., Shirazy, A., Hashim, M., 2022. Neuro-fuzzy-AHP (NFAHP) technique for copper exploration using advanced
spaceborne thermal emission and reflection radiometer (ASTER) and geological datasets in the sahlabad mining area, east Iran. Rem. Sens. 14 (21), 5562.
Shirazy, A., Ziaii, M., Hezarkhani, A., Timkin, T., 2020. Geostatistical and remote sensing studies to identify high metallogenic potential regions in the Kivi area of Iran:
minerals. 10 (10), 869.
Shirmard, H., Farahbakhsh, E., Heidari, E., Beiranvand Pour, A., Pradhan, B., Müller, D., Chandra, R., 2022. A comparative study of convolutional neural networks and
conventional machine learning models for lithological mapping using remote sensing data. Rem. Sens. 14 (4), 819.
Siebels, K., Goïta, K., Germain, M., 2020. Estimation of mineral abundance from hyperspectral data using a new supervised neighbor-band ratio unmixing approach.
IEEE Trans. Geosci. Rem. Sens. 58 (10), 6754–6766.

27
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

Singer, R.B., McCord, T.B., 1979. Mars-large scale mixing of bright and dark surface materials and implications for analysis of spectral reflectance. 2.(A80-23617 08-
91). In: Proceedings in: Lunar and Planetary Science Conference, 10th, Houston, Tex., March 19-23, 1979, Proceedings, vol. 10. Pergamon Press, Inc., New York,
pp. 1835–1848. 1835-1848.1979.
Souza, M.V., Horbe, A.M.C., da Silva, B.C., Peixoto, S.F., Castro, R.T., 2021. Regolith LANDSAT-8/OLI and Hyperion/EO-1 images classification in midwest of Brazil. J.
S. Am. Earth Sci. 111, 103460.
Sun, L., Neville, R., Staenz, K., White, H.P., 2008. Automatic destriping of Hyperion imagery based on spectral moment matching. Can. J. Rem. Sens. 34 (Suppl. 1),
S68–S81.
Sun, W., Du, Q., 2019. Hyperspectral band selection: a review. IEEE Geoscience and Remote Sensing Magazine 7 (2), 118–139.
Sun, W., Liu, S., Wang, M., Zhang, X., Shang, K., Liu, Q., 2023. Soil copper concentration map in mining area generated from AHSI remote sensing imagery. Sci. Total
Environ. 860, 160511.
Tamilarasan, K., Anbazhagan, S., Maheswaran, S.U., Ranjithkumar, S., Kusumab, K., Rajeshc, V., 2022. Reflectance spectra and AVIRIS-NG airborne hyperspectral data
analysis for mapping ultramafic rocks in igneous terrain. J. Spectr. Imaging 11.
Tan, Y., Lu, L., Bruzzone, L., Guan, R., Chang, Z., Yang, C., 2020. Hyperspectral band selection for lithologic discrimination and geological mapping. IEEE J. Sel. Top.
Appl. Earth Obs. Rem. Sens. 13, 471–486.
Tao, X., Paoletti, M.E., Han, L., Wu, Z., Ren, P., Plaza, J., Plaza, A., Haut, J.M., 2022. A new deep convolutional network for effective hyperspectral unmixing. IEEE J.
Sel. Top. Appl. Earth Obs. Rem. Sens. 15, 6999–7012.
Thannoun, R.G., 2021. Mapping lithological and mineralogical units using hyperspectral imagery: Malaysian. J. Sci. 93–106.
Thiele, S.T., Bnoulkacem, Z., Lorenz, S., Bordenave, A., Menegoni, N., Madriz, Y., Dujoncquoy, E., Gloaguen, R., Kenter, J., 2021. Mineralogical mapping with
accurately corrected shortwave infrared hyperspectral data acquired obliquely from UAVs. Rem. Sens. 14 (1), 5.
Transon, J., d’Andrimont, R., Maugnard, A., Defourny, P., 2018. Survey of hyperspectral earth observation applications from space in the sentinel-2 context. Rem. Sens.
10 (2), 157.
Tripathi, M.K., Govil, H., 2019a. Evaluation of AVIRIS-NG hyperspectral images for mineral identification and mapping. Heliyon 5 (11), e02931.
Tripathi, M.K., Govil, H., 2019b. Evaluation of AVIRIS-NG hyperspectral images for mineral identification and mapping. Heliyo 5 (11).
Tripathi, M.K., Govil, H., Chattoraj, S., 2020. Identification of hydrothermal altered/weathered and clay minerals through airborne AVIRIS-NG hyperspectral data in
Jahajpur, India. Heliyo 6 (2).
Tripathi, P., Garg, R.D., 2021. Feature extraction of DESIS and PRISMA hyperspectral remote sensing datasets for geological applications: the International Archives of
the Photogrammetry. Remote Sensing and Spatial Information Sciences 44, 169–173.
Tripathi, P., Garg, R.D., 2023. Potential of DESIS and PRISMA hyperspectral remote sensing data in rock classification and mineral identification: a case study for
Banswara in Rajasthan, India. Environ. Monit. Assess. 195 (5), 575.
Van Der Meer, F., 2000. Geophysical inversion of imaging spectrometer data for geologic modelling. Int. J. Rem. Sens. 21 (2), 387–393.
Van der Meer, F.D., Van der Werff, H.M., Van Ruitenbeek, F.J., Hecker, C.A., Bakker, W.H., Noomen, M.F., Van Der Meijde, M., Carranza, E.J.M., De Smeth, J.B.,
Woldai, T., 2012. Multi-and hyperspectral geologic remote sensing: a review. Int. J. Appl. Earth Obs. Geoinf. 14 (1), 112–128.
Veganzones, M.A., Grana, M., 2008. Endmember extraction methods: a short review. In: Proceedings International Conference on Knowledge-Based and Intelligent
Information and Engineering Systems. Springer, pp. 400–407.
Wambugu, N., Chen, Y., Xiao, Z., Tan, K., Wei, M., Liu, X., Li, J., 2021. Hyperspectral image classification on insufficient-sample and feature learning using deep neural
networks: a review. Int. J. Appl. Earth Obs. Geoinf. 105, 102603.
Wan, Y.-q., Fan, Y.-h., Jin, M.-s., 2021. Application of hyperspectral remote sensing for supplementary investigation of polymetallic deposits in Huaniushan ore region,
northwestern China. Sci. Rep. 11 (1), 440.
Wang, F., Yi, Q., Hu, J., Xie, L., Yao, X., Xu, T., Zheng, J., 2021a. Combining spectral and textural information in UAV hyperspectral images to estimate rice grain yield.
Int. J. Appl. Earth Obs. Geoinf. 102, 102397.
Wang, J., Chang, C.-I., 2006. Independent component analysis-based dimensionality reduction with applications in hyperspectral image analysis. IEEE Trans. Geosci.
Rem. Sens. 44 (6), 1586–1600.
Wang, M., Huang, Z., Zhang, X., Zhang, Y., Chen, M., 2021b. Altered mineral mapping based on ground-airborne hyperspectral data and wavelet spectral angle mapper
tri-training model: case studies from Dehua-Youxi-Yongtai Ore District, Central Fujian, China. Int. J. Appl. Earth Obs. Geoinf. 102, 102409.
Wang, S., Zhou, K., Wang, J., Zhao, J., 2022. Identifying and mapping alteration minerals using HySpex airborne hyperspectral data and random forest algorithm. Front.
Earth Sci. 10, 871529.
Wang, Z., Zuo, R., Jing, L., 2021c. Fusion of geochemical and remote-sensing data for lithological mapping using random forest metric learning. Math. Geosci. 53,
1125–1145.
Wang, Z., Zuo, R., Liu, H., 2021d. Lithological mapping based on fully convolutional network and multi-source geological data. Rem. Sens. 13 (23), 4860.
Wei, J., Wang, X., 2020. An overview on linear unmixing of hyperspectral data. Math. Probl Eng. 2020, 1–12.
Wu, C., Dai, J., Zhou, A., He, L., Tian, B., Lin, W., Zhao, T., Bai, L., 2023. Mapping alteration zones in the Southern section of Yulong copper belt, Tibet using multi-
source remote sensing data. Front. Earth Sci. 11, 1164131.
Wu, N., 2012. The Maximum Entropy Method. Springer Science & Business Media.
Xia, J., Du, P., He, X., Chanussot, J., 2013. Hyperspectral remote sensing image classification based on rotation forest. Geosci. Rem. Sens. Lett. IEEE 11 (1), 239–243.
Xu, C., Fu, L., Lin, T., Li, W., Ma, S., 2022. Machine Learning in Petrophysics: Advantages and Limitations: Artificial Intelligence in Geosciences, vol. 3. pp. 157–161.
Xu, Q., Liu, S., Ye, F., Zhang, Z., Zhang, C., 2018. Application of CASI/SASI and fieldspec4 hyperspectral data in exploration of the Baiyanghe uranium deposit,
Hebukesaier, Xinjiang, NW China. Int. J. Rem. Sens. 39 (2), 453–469.
Xu, Y., Li, Z., Xie, Z., Cai, H., Niu, P., Liu, H., 2021. Mineral prospectivity mapping by deep learning method in Yawan-Daqiao area, Gansu. Ore Geol. Rev. 138, 104316.
Yang, X., Chen, J., Wang, C., Chen, Z., 2022. Residual dense autoencoder network for nonlinear hyperspectral unmixing. IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens.
15, 5580–5595.
Ye, B., Tian, S., Cheng, Q., Ge, Y., 2020. Application of lithological mapping based on advanced hyperspectral imager (AHSI) imagery onboard Gaofen-5 (GF-5) satellite.
Rem. Sens. 12 (23), 3990.
Yu, F., Koltun, V., 2015. Multi-scale Context Aggregation by Dilated Convolutions. arXiv preprint arXiv:1511.07122.
Yu, J., Zhang, L., Li, Q., Li, Y., Huang, W., Sun, Z., Ma, Y., He, P., 2021. 3D autoencoder algorithm for lithological mapping using ZY-1 02D hyperspectral imagery: a case
study of Liuyuan region. J. Appl. Remote Sens. 15 (4), 42610. 042610.
Yu, S., Jia, S., Xu, C., 2017. Convolutional neural networks for hyperspectral image classification. Neurocomputing 219, 88–98.
Zadeh, L.A., 1965. Fuzzy Sets, Fuzzy Sets, Fuzzy Logic, and Fuzzy Systems: Selected Papers by Lotfi A Zadeh. World Scientific, pp. 394–432.
Zhang, C., Yi, M., Ye, F., Xu, Q., Li, X., Gan, Q., 2022. Application and evaluation of deep neural networks for airborne hyperspectral remote sensing mineral mapping: a
case study of the Baiyanghe uranium deposit in northwestern Xinjiang, China. Rem. Sens. 14 (20), 5122.
Zhang, L., Zhang, M., Huang, J., Zhang, C., Ye, F., Pan, W., 2023. A new approach for mineral mapping using drill-core hyperspectral image. Geosci. Rem. Sens. Lett.
IEEE 20, 1–5.
Zhang, S., Xiao, K., Carranza, E.J.M., Yang, F., 2019. Maximum entropy and random forest modeling of mineral potential: analysis of gold prospectivity in the
Hezuo–Meiwu district, west Qinling Orogen, China. Nat. Resour. Res. 28, 645–664.
Zhao, H., Deng, K., Li, N., Wang, Z., Wei, W., 2020. Hierarchical spatial-spectral feature extraction with long short term memory (LSTM) for mineral identification using
hyperspectral imagery. Sensors 20 (23), 6854.
Zheng, P., Wu, Z., Sun, J., Zhang, Y., Zhu, Y., Shen, Y., Yang, J., Wei, Z., Plaza, A., 2021. A parallel unmixing-based content retrieval system for distributed
hyperspectral imagery repository on cloud computing platforms. Rem. Sens. 13 (2), 176.
Zheng, P., Wu, Z., Zhang, W., Li, M., Yang, J., Zhang, Y., Wei, Z., 2018. An unmixing-based content retrieval method for hyperspectral imagery repository on cloud
computing platform. In: Proceedings IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium. IEEE, pp. 3583–3586.
Zhizhong, L., Rihong, Y., Fuxing, D., Peijun, D., Xianfeng, Z., Bingxiang, T., Huijie, Z., Hongjun, S., 2012. A review on the geological applications of hyperspectral

28
S. Hajaj et al. Remote Sensing Applications: Society and Environment 35 (2024) 101218

remote sensing technology. In: Proceedings 2012 4th Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS). IEEE,
pp. 1–4.
Zhu, J., Wu, L., Hao, H., Song, X., Lu, Y., 2017. Auto-encoder based for high spectral dimensional data classification and visualization. In: Proceedings 2017 IEEE Second
International Conference on Data Science in Cyberspace (DSC). IEEE, pp. 350–354.
Zhu, L., Chen, Y., Ghamisi, P., Benediktsson, J.A., 2018. Generative adversarial networks for hyperspectral image classification. IEEE Trans. Geosci. Rem. Sens. 56 (9),
5046–5063.
Zinno, I., Elefante, S., Mossucca, L., De Luca, C., Manunta, M., Terzo, O., Lanari, R., Casu, F., 2015. A first assessment of the P-SBAS DInSAR algorithm performances
within a cloud computing environment. IEEE J. Sel. Top. Appl. Earth Obs. Rem. Sens. 8 (10), 4675–4686.
Zortea, M., Plaza, A., 2009. Spatial preprocessing for endmember extraction. IEEE Trans. Geosci. Rem. Sens. 47 (8), 2679–2693.

29