Review

Artificial Intelligence in Agro-Food Systems: From Farm to Fork

Ali Aghababaei 1 , Fatemeh Aghababaei 2 , Marc Pignitter 3 and Milad Hadidi 3, *

1 Department of Information Engineering, University of Padova, Via Gradenigo, 6/b, 35131 Padova, Italy;
[email protected]
2 Aora Health, Calle Via de los Poblados, 17, 28033 Madrid, Spain; [email protected]
3 Institute of Physiological Chemistry, Faculty of Chemistry, University of Vienna, 1090 Vienna, Austria;
[email protected]
* Correspondence: [email protected] or [email protected]

Abstract: The current landscape of the food processing industry places a strong emphasis
on improving food quality, nutritional value, and processing techniques. This focus arises
from consumer demand for products that adhere to high standards of quality, sensory
characteristics, and extended shelf life. The emergence of artificial intelligence (AI) and
machine learning (ML) technologies is instrumental in addressing the challenges associated
with variability in food processing. AI represents a promising interdisciplinary approach
for enhancing performance across various sectors of the food industry. Significant ad-
vancements have been made to address challenges and facilitate growth within the food
sector. This review highlights the applications of AI in agriculture and various sectors
of the food industry, including bakery, beverage, dairy, food safety, fruit and vegetable
industries, packaging and sorting, and the drying of fresh foods. Various strategies have
been implemented across different food sectors to promote advancements in technology.
Additionally, this article explores the potential for advancing 3D printing technology to
enhance various aspects of the food industry, from manufacturing to service, while also
outlining future perspectives.

Keywords: machine learning; food industry; agriculture sector; computer science; artificial
neural networks; algorithms

Academic Editors: Panagiotis

Tsakanikas and Guangnan Chen 1. Introduction
Received: 10 December 2024
Artificial intelligence (AI) is a branch of computer science focused on replicating
Revised: 28 December 2024
human cognitive processes, including reasoning, learning, and knowledge retention [1].
Accepted: 24 January 2025
Published: 27 January 2025
AI is divided into two categories: strong AI and weak AI. Weak AI is designed to create
machines that simulate human behavior and judgments, whereas strong AI posits that
Citation: Aghababaei, A.;
Aghababaei, F.; Pignitter, M.; Hadidi,
machines can truly replicate human cognitive processes and consciousness [2]. Strong AI
M. Artificial Intelligence in Agro-Food has not yet been realized, and research in this area is ongoing. AI methods are currently
Systems: From Farm to Fork. Foods being applied across various fields, including the gaming industry, weather forecasting,
2025, 14, 411. https://doi.org/ heavy and process industries, the food and medical sectors, data mining, stem cell research,
10.3390/foods14030411
and knowledge representation [3–5]. AI offers a wide range of algorithms, including
Copyright: © 2025 by the authors. reinforcement learning, expert systems (ESs), fuzzy logic (FL), swarm intelligence, the
Licensee MDPI, Basel, Switzerland. Turing test, cognitive science, artificial neural networks (ANNs), and logic programming.
This article is an open access article
The impressive performance of AI has made it a highly favored tool in industries for
distributed under the terms and
decision-making and process estimation, with the goal of reducing costs, enhancing quality,
conditions of the Creative Commons
Attribution (CC BY) license
and improving profitability [6].
(https://creativecommons.org/ With the global population projected to increase, food demand is expected to grow
licenses/by/4.0/). by 59 to 98% by 2050. To meet this rising demand, AI is being utilized in various areas,

Foods 2025, 14, 411 https://doi.org/10.3390/foods14030411

Foods 2025, 14, 411 2 of 38

including supply chain management, food sorting, production optimization, quality en-
hancement, and maintaining industrial hygiene standards [7]. According to Sharma, the
food processing and handling industries are projected to experience a compound annual
growth rate of approximately 5% at least through 2021 [7]. According to Funes and col-
leagues, ANNs have been employed as an effective tool for solving complex real-world
problems in the food industry [8]. Correa et al. [9] further noted that ANNs simplify
the classification and prediction of parameters, contributing to their increased adoption
in recent years. Additionally, FL and ANNs have been utilized as controllers to ensure
food safety, quality control, yield improvement, and cost reduction in production [10]. AI
technologies have also proven valuable in food drying processes, serving as effective tools
for process control in this area [11–13].
Previous research has demonstrated numerous applications of AI in the food industry,
each targeting specific goals. One study focused on the various applications of ANN in food
process modeling, though it primarily emphasized the use of ANNs within this particular
area [14]. Additionally, the application of AI technologies such as ANNs, FL, and ESs in
the food industry has been reviewed, with a particular emphasis on the drying of fresh
fruits [11]. A review has examined how food safety remains a primary concern within the
food industry, prompting the development of smart packaging systems to meet the needs of
the food supply chain. These intelligent packaging systems monitor the condition of food,
providing information about its quality throughout storage and transportation. The review
on intelligent packaging as a tool to reduce food waste reported approximately 45 recent
advancements in optical systems for monitoring freshness. The study focused on meat, fish
products, fruits, and vegetables, as these are the most common areas of application [15].
Several studies have been conducted on intelligent packaging, demonstrating that its use
plays a crucial role in the food industry. These systems are capable of monitoring the
freshness of food products and crops throughout the food supply chain [11,16,17].
Although several studies have explored the use of AI and sensors in the food industry,
their scope remains limited. Consequently, a comprehensive review that consolidates all AI
applications in the food industry, along with their integration with suitable sensors, would
be highly beneficial. To the author’s knowledge, such a review is currently unavailable. This
type of resource would provide a valuable one-stop reference for industry professionals,
practitioners, and academics, detailing the advantages, limitations, and methodologies of
these technologies. Specifically, various types of AI and their recent applications in the food
industry will be highlighted, covering several AI techniques such as ESs, FL, ANNs, and
ML. In a later section, a critical review is conducted that discusses the primary applications
of AI algorithms in the food industry. Following this, the trends in AI applications within
the food industry are illustrated.

2. Scope and Approach

Food production operates within a complex network of supply chain participants
that interact on a global scale. The efficiency of this system is shaped by interconnected
international and local factors. Numerous key drivers have been identified as influencing
the food supply chain, including (i) the expanding global population and evolving dietary
habits, (ii) resource shortages essential for agricultural production, such as fertile land, fresh
water, and energy, (iii) climate change, (iv) declining biodiversity, (v) poor governance, and
(vi) competing agricultural systems [18].
The current food system is inefficient in addressing these challenges and continues to
be excessively wasteful, polluting, and demanding on resources. If these trends continue,
it will not be able to provide a healthy, safe, and nutritious diet to a growing global
population while remaining within the planet’s ecological boundaries. A shift towards a
Foods 2025, 14, 411 3 of 38

sustainable food system that can meet these needs necessitates adopting a holistic systems
approach [19].
In a systems approach, all elements are in constant interaction, often through feedback
loops within the food supply chain. This method is particularly crucial for complex sys-
tems like food supply chains, where many interactions occur between various influencing
factors through mechanisms that are still not completely understood. Various conceptual
approaches to food systems have been created, integrating all activities and their inter-
connections within the system. These approaches also consider food security alongside
socio-economic and environmental indicators [20]. These approaches take a holistic view
by broadening the traditional understanding of the food supply chain or farming system to
include the interactions within the whole food system and its socio-economic and biophysi-
cal environment. Such frameworks enable the development of comprehensive international
policies that target food security, nutrition, and global agribusiness. Achieving a truly
sustainable food system requires aligning the digital strategy vision with the objectives
of the Green Deal. In this framework, we outline our vision for a systems approach to
food production, highlighting how digital technologies and AI can assist in overcoming
the challenges associated with it.

3. Artificial Intelligence: State-of-the-Art Techniques

DL and machine learning (ML) are two highly utilized AI techniques across individual,
business, and governmental sectors, leveraging data-driven models to enable accurate
predictions. In the food industry, recent developments in ML aim to manage the complexi-
ties and uncertainties of vast information [21]. Core subfields within AI, such as ANNs,
robotics, ESs, computer vision systems (CVSs), natural language processing (NLP), and
ML, each contribute distinct capabilities. NLP aids in interpreting human language, CVSs
facilitate digital-to-analog conversion (e.g., speech recognition and video analysis), and ESs
replicate human-like decision-making. The three cognitive functions underlying AI are
learning (data acquisition and algorithm development for deriving actionable information),
reasoning, and self-correction [22].
Advances in AI have recently expanded its use within the agro-food sector, where it
plays a critical role in identifying models, generating services, and supporting decision-
making. AI contributes substantially to agricultural productivity by providing accurate,
predictive insights for optimizing resource usage [23]. Specifically, AI algorithms classify
patterns, predict potential issues, and aid in managing agricultural challenges, such as
pest identification and treatment recommendations, irrigation, and water management via
smart irrigation systems. Remote sensing and sensor technology assess biotic and abiotic
factors, enhancing agricultural and livestock management practices [22].
The data acquisition process gathers real-world information from sources such
as sensors, historical records, and customer feedback. This is followed by data
pre-processing—cleaning, normalizing, and standardizing the data to ensure its consis-
tency and quality. Next, pertinent features are extracted from the data through feature
engineering to develop an AI model capable of accurately predicting output quality. The
pre-processed data then undergo model training, allowing for accurate product quality
predictions. Throughout this process, human oversight plays a vital role in tasks such as
data annotation, model validation, interpretability, and ethical compliance, ensuring that
the AI model functions effectively and adheres to quality standards and regulations [22].

3.1. Machine Learning

ML is a pivotal AI field driving the advancement of creative and productive work.
As illustrated in Figure 1A, ML can be categorized into two primary types—supervised
 3.1. Machine Learning
Foods 2025, 14, 411 ML is a pivotal AI field driving the advancement of creative and productive work.4 of 38
As illustrated in Figure 1A, ML can be categorized into two primary types—supervised
and unsupervised learning—each encompassing various methodologies applicable across
and unsupervised
diverse domains, learning—each encompassing
including food processing, variousinmethodologies
as depicted applicable
Figure 1B. The principal across
ML
diverse domains,
tasks are including
supervised food processing,
and unsupervised learningas[24].
depicted in Figure 1B. The principal ML
tasks are supervised and unsupervised learning [24].

Figure 1. 1.(A)
Figure (A)classification
classification of ML and
of ML and(B)
(B)ML
MLinterpretation
interpretationin in
thethe food
food business
business using
using different
different
algorithms.
algorithms.

InIn
supervised
supervisedlearning,
learning,the
theprocess
process requires
requires labeled data and
labeled data andoversight.
oversight.Here,
Here,com-
comput-
ersputers
are trained on a “labeled” dataset, allowing them to accurately map input
are trained on a “labeled” dataset, allowing them to accurately map input to outputto output
variables. After training,
variables. After training, the model is able to make predictions based on these
model is able to make predictions based on these mappings. mappings.
Applications
Applicationsofofsupervised
supervisedlearning
learning span
span risk assessment,
assessment,spam
spamfiltering,
filtering,and
and fraud
fraud de-detec-
tection.
tion. Conversely,
Conversely, unsupervised
unsupervised learning
learning enables
enables computers
computers to analyze
to analyze data data inde-
independently
pendently
using usingdatasets
unlabeled unlabeled datasets
[25]. [25]. Thisinvolves
This method method grouping
involves grouping or categorizing
or categorizing data based
on patterns, similarities, or differences, with the aim of discovering hidden structures
within the dataset. ML leverages mathematical and statistical methods to draw insights
and make decisions from data. Broadly, ML techniques fall into symbolic and sub-symbolic
approaches. In supervised learning, for instance, the goal is to create a predictive model
that maps input variables to a specified output variable using labeled data [26]. Among
the algorithms frequently utilized in supervised learning are decision trees (DTs), Bayesian
networks (BNs), and regression analysis.
Foods 2025, 14, 411 5 of 38

In contrast, unsupervised learning operates on unlabeled data to uncover hidden

patterns, primarily applied in tasks like dimensionality reduction and exploratory data
analysis. This approach combines training and test data, allowing the learner to engage
with its environment to gather insights. Through exploratory actions, rather than relying
solely on pre-existing data, the learner enhances its understanding by experimenting with
new, untested actions [27].

3.1.1. Deep Learning

DL is a subset of ML that involves training ANNs with multiple layers to model
complex patterns in data. This approach has been particularly successful in tasks such
as image and speech recognition, natural language processing, and game playing [28].
The development of DL has been significantly influenced by researchers like Yann LeCun,
Yoshua Bengio, and Geoffrey Hinton, who have contributed foundational work in this
field. For instance, LeCun’s work on convolutional networks has been pivotal in advancing
image recognition technologies. The success of deep learning is attributed to its ability to
automatically learn hierarchical representations from raw data, enabling the modeling of
DL to form a cornerstone of modern AI research.
DL recently experienced significant advancements in the detection of food adulteration
and defects. This progress began in the late 1990s with the development of convolutional
neural networks (CNNs), which have become one of the most powerful methods for learn-
ing intricate features of digital data in tasks related to classification and regression [29].
Unlike traditional neural networks, CNNs employ multiple convolutional layers, as il-
lustrated in the standard CNN architecture for food analysis and detection in Figure 2A.
These convolutional layers utilize filters to extract features from input images, with specific
parameters—such as padding, stride, kernel size, and activation functions—tailored to the
task at hand and optimized for effectiveness.
At the network’s output end, fully connected layers act as classifiers, utilizing a densely
connected block to generate the final prediction. Various CNN architectures have been
developed for image classification in food and agricultural applications, including LeNet,
AlexNet, VGG, GoogLeNet, ResNet, and Inception. CNNs have proven highly effective in
Foods 2025, 14, x FOR PEER REVIEWidentifying food adulteration and evaluating the quality of agricultural products. However, 6 of 38
GoogLeNet, ResNet, and Inception do not specifically detail their performance in food and
agriculture applications [30].

Figure 2. Cont.
Foods 2025, 14, 411 6 of 38

Figure 2. A schematic representation of (A) a CNN and (B) an ANN.

Figure 2. A schematic representation of (A) a CNN and (B) an ANN.
3.1.2. Artificial Neural Network
3.1.2.An
Artificial Neural Network
ANN represents a sophisticated component of AI with substantial potential for
An ANN
application withinrepresents
the foodaindustry.
sophisticated component
Particularly of AI withlearning,
in unsupervised substantial
ANNspotential for
leverage
unlabeled
applicationdatasets,
within functioning withoutParticularly
the food industry. predefined in input and outputlearning,
unsupervised variables.ANNs
Techniques
lever-
such as ANNs, datasets,
age unlabeled clustering, genetic algorithms,
functioning and DL play
without predefined inputa vital
androle,
outputwith ANNs being
variables. Tech-
the mostsuch
niques prominent
as ANNs, approach in genetic
clustering, evaluating food quality
algorithms, and DL within
play AI-based applications.
a vital role, with ANNs
Structurally,
being the most an ANN is a network
prominent approachofinnodes that can
evaluating perform
food qualityeither
within linear or nonlinear
AI-based applica-
processing tasks [31].an
tions. Structurally, Theoretically, electricalofimpulses
ANN is a network nodes that pass through
can perform synaptic
eitherconnections
linear or non-at
each
linearnode within the
processing interconnected
tasks network,
[31]. Theoretically, movingimpulses
electrical through neuron-like
pass through structures
synapticwith
con-
the aid of at
nections axons
each[32].
node ANNs
withinoperate on principles similar
the interconnected network, to moving
the human brain,neuron-like
through triggering
internal operations
structures with therather
aid ofthan
axonsconventional
[32]. ANNscomputational processes.
operate on principles similar to the human
A triggering
brain, standard ANN architecture
internal operations comprises
rather thanthree main layers:computational
conventional input, hidden, processes.
and output
(as illustrated
A standard in Figure 2B). The network’s
ANN architecture comprises architecture
three mainincorporates
layers: input, activation
hidden, andfunctions,
output
with data processed
(as illustrated through
in Figure feed-forward
2B). The network’s or feedback mechanisms.
architecture The inputfunctions,
incorporates activation layer ini-
tially
with receives raw data,
data processed which then
through flows to the
feed-forward or hidden
feedback layer for processing
mechanisms. Thebefore
input reaching
layer ini-
the output
tially layer
receives rawthrough a sequence
data, which of interconnected
then flows to the hidden layer nodes. forOne major advantage
processing before reach-of
ANN
ing thelayers
outputlieslayer
in their ability
through to facilitate
a sequence parallel reasoning,
of interconnected nodes. making neural
One major networks
advantage of
highly effective
ANN layers liesfor
in prediction.
their ability Similar to a human
to facilitate parallel brain, an ANN
reasoning, makingcan neural
learn and store
networks
synaptic weights,for
highly effective representing
prediction.connections
Similar to between
a humanneuronsbrain, an [22].
ANN can learn and store
According
synaptic weights, to Sukhadia
representing andconnections
Chaudhari between
[33], ANN structures
neurons [22].are tailored to support
specific applications, such as pattern recognition and data classification. Furthermore,
Gonzalez-Fernandez et al. [34] emphasize that ANNs are adaptable, versatile, and suitable
for various scenarios. Although some modifications may be necessary, ANNs are highly
flexible and can simulate a wide range of nonlinear systems. Nonlinear regression is among
the network’s most distinct capabilities. The multilayer perceptron (MLP) model, frequently
used for pattern recognition and predictive tasks, processes input data in its hidden layer
to form internal representations. Throughout training, the network’s parameters—weights
and biases—interact in complex ways to develop these representations. While the opaque,
“black-box” nature of ANNs often challenges precise interpretation, tools such as feature
analysis, visualization, and weight interpretation offer insights into the patterns recognized
by neurons within the hidden layer [22].
 eters—weights and biases—interact in complex ways to develop these representa
While the opaque, “black-box” nature of ANNs often challenges precise interpre
tools such as feature analysis, visualization, and weight interpretation offer insigh
Foods 2025, 14, 411 the patterns recognized by neurons within the hidden layer [22]. 7 of 38

3.2. Computer Vision System

3.2. Computer Vision System
Lukinac et al. [35] describe a series of procedures in CVSs, including imag
Lukinac etcessing,
al. [35] describe a seriesanalysis,
digitalization, of procedures in CVSs,
and capture including
(Figure image
3). This processing,
method relies on the
digitalization, analysis,
spectrumand capture (Figure
interacting with a 3). This method
reflective relies onmaterial,
or absorbent the visible spectrum
capturing images th
interacting with a reflective or absorbent material, capturing images through image
image analysis. During this process, photons are captured by camera lenses and the analy-
sis. During thisverted
process,
intophotons
electricalaresignals
capturedby anbyimage
camera lensesDigitization
sensor. and then converted into
involves transforming
electrical signals by an image sensor. Digitization involves transforming
images into a numerical format for further analysis. They observed that these images into
a CVS can
a numerical format for further
tify external analysis.
attributes They observed
in digital that a CVSautomated
images, facilitating can quantify external
quality control of pro
attributes in digital
[35]. images, facilitating
As AI continues to automated quality control
progress, integrating ML, of products
ESs, FL, and [35]. As offers
CVSs AI subs
continues to progress, integrating
opportunities ML, ESs, FL,
for innovation and multiple
across CVSs offers substantial
fields. opportunities
For complex, multi-layered pr
for innovation across
tion and multiple fields. For
consumption complex,
networks, multi-layered
optimizing production
AI models and consump-
is essential to increase effic
tion networks,adaptability,
optimizing AI andmodels is essential
resilience. to increase
Streamlined efficiency,
AI models adaptability,
can reduce resourceandwaste, lowe
resilience. Streamlined AI models can reduce resource waste, lower production costs,
duction costs, and enhance productivity. Effective resource allocation is crucial in th and
enhance productivity.
timization,Effective resource allocation
with algorithms designed to is crucial
ensure in this optimization,
resources withacross ne
are distributed
algorithms designed
nodes,tominimizing
ensure resources
shortagesare distributed
or excess. across network nodes, minimizing
shortages or excess.

Figure 3. Schematic representation of CVS.

Figure 3. Schematic representation of CVS.

AI-driven supply chain management also focuses on precise demand forecasting,

AI-driven supply chain management also focuses on precise demand forecasti
inventory optimization, and disruption identification, all contributing to uninterrupted
ventory optimization, and disruption identification, all contributing to uninterrupt
operations. Additionally, AI can integrate risk mitigation measures, identifying and ad-
erations. Additionally, AI can integrate risk mitigation measures, identifying and ad
dressing potential risks, such as supply chain disturbances or market volatility. AI models
ing potential risks, such as supply chain disturbances or market volatility. AI mode
continuously adapt to shifting conditions, enhancing resilience and maintaining perfor-
tinuously adapt to shifting conditions, enhancing resilience and maintaining perfor
mance in dynamic environments. Through these optimization strategies, AI contributes to
greater efficiency, flexibility, and reliability within complex networks, helping to meet the
demands of rapidly changing operational landscapes [22].

3.3. Fuzzy Logic

FL effectively simulates the complex decision-making processes humans use when
faced with ambiguous or uncertain information. It is recognized as a valuable ap-
proach for managing classification challenges involving incomplete data. However, FL
requires enhancement in handling recovery processes, especially under constraints in
high-dimensional or large datasets. Fuzzy systems facilitate simplified algorithmic formu-
lations and linguistic variables to represent the behavior of complex systems qualitatively
rather than relying on quantitative or mathematical models. FL has been instrumental in
developing control systems for sophisticated food and beverage manufacturing processes.
 enhancement in handling recovery processes, especially under constraints in high-dimen-
sional or large datasets. Fuzzy systems facilitate simplified algorithmic formulations and
Foods 2025, 14, 411 linguistic variables to represent the behavior of complex systems qualitatively rather8than of 38
relying on quantitative or mathematical models. FL has been instrumental in developing
control systems for sophisticated food and beverage manufacturing processes.
A typical FL architecture, as illustrated
illustrated in
in Figure
Figure 4,
4, comprises
comprises four
four essential
essential compo-
compo-
nents: the
the rule base, fuzzification, inference engine, and defuzzification.
defuzzification. According
According to
fuzzy set theory, an element is part of a fuzzy set if it is associated with a real number
between 0 and 11 [36].
[36]. FL
FLmodels
modelsinvolve
involvevarious
variousstages,
stages,including
includingfuzzification,
fuzzification,inference,
inference,
and defuzzification
defuzzification [37]. Fuzzification
Fuzzification transforms
transforms aa crisp
crisp input
input value
value into
into aa degree of
membership
membership within
within aa fuzzy
fuzzy set, generally
generally represented
represented on a scale from 0 to 1. The most
commonly used
used membership
membershipfunctions
functionsinclude
includetriangular,
triangular, Z-shaped,
Z-shaped, S-shaped,
S-shaped, trapezoi-
trapezoidal,
dal,
and and Gaussian
Gaussian forms,
forms, though
though numerous
numerous otherother
typestypes
existexist
[38]. [38].

Figure 4. The main four parts of FL.

The inference
The inference system
systemisiswhere
wherefuzzy fuzzyrules areare
rules applied
appliedto translate fuzzy
to translate inputs
fuzzy intointo
inputs cor-
responding outputs, leading to the final defuzzification stage [39]. Defuzzification
corresponding outputs, leading to the final defuzzification stage [39]. Defuzzification methods
vary, withvary,
methods commonwithtechniques including the
common techniques mean ofthe
including maximum,
mean of center
maximum,of maximum,
center ofcenter
max-
of gravity, centroid of area, smallest of maximum, and largest of maximum.
imum, center of gravity, centroid of area, smallest of maximum, and largest of maximum. Villasenor-
Aguilar et al. [40] applied
Villasenor-Aguilar FL applied
et al. [40] to predict FLtotal solubletotal
to predict solids and assess
soluble solidsbell
andpepper maturity
assess bell pep-
stages, reaching an 88% classification precision for the four stages of maturity.
per maturity stages, reaching an 88% classification precision for the four stages of ma- Additionally,
Pakyürek
turity. et al. [41] utilized
Additionally, Pakyürek FL et foral.quality grading
[41] utilized FLacross three grading
for quality pineapple varieties.
across three pine-
These studies
apple varieties. highlight FL’s robustness as a control mechanism, particularly in han-
dlingThese
intricate processes essential for assessing food and agricultural product
studies highlight FL’s robustness as a control mechanism, particularly in han- quality.
dling intricate processes essential for assessing food and agricultural product quality.
3.4. Expert System
An ES,
3.4. Expert a subset of AI, leverages specialized knowledge and reasoning capabilities
System
to tackle complex problems through structured knowledge representation. Typically, an
An ES, a subset of AI, leverages specialized knowledge and reasoning capabilities to
ES comprises an interpreter, an inference engine, a dynamic database, a human–machine
tackle complex problems through structured knowledge representation. Typically, an ES
interface, and a knowledge acquisition module (Figure 5). The problem-solving compo-
comprises an interpreter, an inference engine, a dynamic database, a human–machine
nent of an ES is designed to mimic expert-level reasoning by processing knowledge and
information systematically. Recently, ESs have been implemented on standard devices
with optimized variables and multiple predictive factors, enhancing their performance. By
selecting features based on color, texture, and geometry, ESs have achieved 100% accuracy
in classifying fruit ripeness [22].
Duong et al. [42] created an ES utilizing EfficientNet and MixNet deep neural net-
work (DNN) architectures to identify fruits. These architectures significantly improved
classification accuracy by up to 95% compared to a widely recognized baseline. To ensure
ESs are commercially viable for quality control in food and agriculture, rigorous develop-
ment processes are essential. Evaluating ESs involves assessing their processing speed,
monitoring capabilities, and effectiveness in automated detection tasks. As input features
are readily available, the system’s accuracy and operational efficiency are key metrics for
successful implementation.
 information systematically. Recently, ESs have been implemented on standard devices
with optimized variables and multiple predictive factors, enhancing their performance.
By selecting features based on color, texture, and geometry, ESs have achieved 100% ac-
Foods 2025, 14, 411 9 of 38
curacy in classifying fruit ripeness [22].

Figure 5. Schematic representation of ES.

Figure 5. Schematic representation of ES.
4. Role of AI in Agriculture
The integration
Duong et al. [42]ofcreated
ML algorithms is playing
an ES utilizing an increasingly
EfficientNet critical role
and MixNet deepacross thenet-
neural four
key stages of the agricultural supply chain: preproduction, production,
work (DNN) architectures to identify fruits. These architectures significantly improved processing, and
distribution accuracy
classification [43]. In the
by preproduction phase,toML
up to 95% compared technologies
a widely are particularly
recognized baseline. Tovaluable
ensure
for forecasting crop yields, assessing soil characteristics, and determining irrigation
ESs are commercially viable for quality control in food and agriculture, rigorous develop- needs.
During
ment the production
processes stage, ML
are essential. can be utilized
Evaluating for detecting
ESs involves assessingcroptheir
diseases and predicting
processing speed,
weather patterns, which are essential for optimizing agricultural output.
monitoring capabilities, and effectiveness in automated detection tasks. As input features In the third
phase,
are which
readily focuses
available, theonsystem’s
processing, ML methods
accuracy are primarily
and operational efficiency employed to optimize
are key metrics for
production planning,
successful implementation. ensuring both high-quality and safe products. Additionally, ML
algorithms play a crucial role in the distribution phase, particularly in areas such as storage
4.management,
Role of AItransportation
in Agriculture logistics, and consumer behavior analysis. The preproduction
phase, which is the starting point of the agricultural supply chain, involves the use of
The integration of ML algorithms is playing an increasingly critical role across the
ML for forecasting crop yields, evaluating soil characteristics, and determining irrigation
four key stages of the agricultural supply chain: preproduction, production, processing,
requirements. Numerous studies highlight the significance of crop yield prediction in
and distribution [43]. In the preproduction phase, ML technologies are particularly valu-
enhancing plant management practices. By incorporating input data such as equipment
able for forecasting crop yields, assessing soil characteristics, and determining irrigation
needs, nutrients, and fertilizers, ML-based predictive models are being developed as part
needs. During the production stage, ML can be utilized for detecting crop diseases and
of precision agriculture tools. These models assist stakeholders and farmers in making
predicting weather patterns, which are essential for optimizing agricultural output. In the
informed decisions regarding crop yield forecasting, thereby advancing smart farming
third phase, which focuses on processing, ML methods are primarily employed to opti-
techniques. Recently, a variety of ML algorithms, including BN, regression models, DTs,
mize production planning, ensuring both high-quality and safe products. Additionally,
clustering methods, DL, and ANNs, have been employed to predict crop yields [44–46].
ML algorithms play a crucial role in the distribution phase, particularly in areas such as
In the field of soil property prediction, various ML algorithms have been applied to
storage management, transportation logistics, and consumer behavior analysis. The pre-
analyze and understand soil characteristics. For instance, Morellos et al. [47] employed
production phase, which is the starting point of the agricultural supply chain, involves
the least squares support vector machine (SVM) method to study a dataset of 140 soil
the use of ML for forecasting crop yields, evaluating soil characteristics, and determining
samples. In another study, Kumar et al. [48] introduced an innovative approach called
irrigation requirements. Numerous studies highlight the significance of crop yield predic-
the crop selection method to address challenges in crop selection and enhance the overall
tion in enhancing plant management practices. By incorporating input data such as equip-
net yield of crops throughout the growing season. Furthermore, Ben Ayed et al. [44]
ment needs, nutrients, and fertilizers, ML-based predictive models are being developed
conducted an analysis of 18 different table olive cultivars from around the world. By
examining morphological, biological, and physicochemical traits, alongside utilizing a
BN, they explored how these factors impact tolerance, productivity, and oil content in the
cultivars. The findings revealed that crop tolerance had a significant impact on oil content.
Additionally, irrigation management plays a crucial role in the preproduction phase, as it
greatly affects both the quality and quantity of the crops. To develop more efficient irrigation
systems—determining the optimal timing, location, and amount of irrigation—researchers
Foods 2025, 14, 411 10 of 38

have utilized input data such as soil moisture levels, precipitation, evaporation rates, and
weather forecasts. These data are then fed into ML models for simulation and optimization,
improving decision-making in irrigation practices [49]. Arvind et al. [50] demonstrated
the effectiveness of integrating ML algorithms with technologies like sensors, Zigbee, and
Arduino microcontrollers for drought prediction and mitigation. Similarly, Cruz et al. [51]
applied an ANN using feed-forward and back-propagation methods to optimize water
resource management within smart farming systems. In a more recent study, Choudhary
et al. [52] utilized partial least squares regression alongside other regression algorithms as
part of an AI toolkit, combining them with sensors for data collection and IoT hardware to
boost efficiency and economic viability. The production phase represents the second stage
of the agricultural supply chain, where various critical factors influence crop production.
Key parameters include weather forecasts, such as sunlight, rainfall, and humidity, as
well as crop protection strategies against both biotic stressors (such as pathogens and
weeds) and abiotic stressors (including nutrient and water deficiencies). Additionally, crop
quality management and harvesting practices play essential roles in ensuring successful
production outcomes. A variety of ML algorithms are employed to create effective models
across different aspects of agriculture. For weather prediction, algorithms such as ANNs,
DL, DTs, ensemble learning, and instance-based learning are commonly used [53]. In crop
protection, clustering and regression methods are applied [54], while ANNs, DTs, DL [55],
and instance-based learning [56] are utilized for weed detection. Crop quality management
typically involves clustering and regression algorithms [57], and for harvesting, DNNs,
along with data mining techniques like K-means clustering, K-nearest neighbor, ANNs,
and SVM, are leveraged [58].
In the final horticultural phase, which occurs during the harvest stage after the crops
have ripened, ML algorithms are also used to predict changes in the color of fruits or crops.
Numerous research teams have employed ML techniques to predict the stages of fruit
ripening and maturity. For instance, Gao et al. [59] attained a 98.6% accuracy in classifi-
cation when they utilized hyperspectral datasets along with the AlexNet-CNN model to
categorize strawberries into early-ripening and fully ripe stages. The processing cluster
represents the third phase in the agricultural supply chain. Various processing methods are
employed for agricultural products, including techniques like smoking, heating, cooking,
cooling, milling, and drying. Choosing the optimal combination of parameters during this
processing phase guarantees both high-quality and high-quantity food production, while
also minimizing the overuse of resources. To accomplish this, many food industries have
adopted modern food processing technologies, integrating software algorithms powered
by ML. Some of the commonly utilized ML algorithms include genetic algorithms, ANNs,
clustering techniques, and BNs [60]. Arora and Mangipudi [61] introduced models based
on SVM classifiers and ANNs to detect nitrosamine in red meat samples. Their predictive
modeling results indicated that the DL model achieved the highest accuracy during testing.
Farah et al. [62] combined differential scanning calorimetry with ML techniques such as
gradient boosting machine (GBM), random forest (RF), and MLP to analyze milk charac-
teristics, verify its authenticity, and detect fraudulent activity. The best outcomes were
obtained using GBM and MLP, which were able to accurately classify 100% of the adulter-
ated samples. The distribution cluster represents the final stage of the agricultural supply
chain, acting as the connection between food production, processing, and the consumer.
ML algorithms can be utilized in various areas, including storage, transportation, consumer
analytics, and inventory management. In the storage and transportation phases, popular
algorithms like genetic algorithms, clustering, and regression techniques are frequently
employed. These predictive techniques are developed to improve food quality preservation,
guarantee product safety, and minimize damage by monitoring the product across the entire
Foods 2025, 14, 411 11 of 38

supply chain [44]. In the consumer analytics phase, ML techniques such as DL and ANNs
are applied in food retailing to predict consumer demand, preferences, and purchasing
patterns. For inventory management, genetic algorithms in ML are utilized to forecast daily
demand and assist in avoiding inventory problems [63]. There are numerous examples of
AI-driven technologies in the agri-food industry, including mechatronics and robotics [64],
drones [64], geographic information systems (GIS) [65], blockchain (BC) [66], and satellite-
based guidance systems [64]. Miranda et al. [67] highlighted these technologies as sensing,
smart, and sustainable, offering systematic processes characterized by connectivity, automa-
tion, precision, monitoring, and digitization [68–70]. In agriculture, robotics, mechatronics,
and smart mechanization are designed to lessen manual labor and optimize resource use by
employing highly autonomous and intelligent machinery [64]. The transition from horses
to tractors, robots, and intelligent vehicles marks a revolutionary period for agriculture and
the food industry. This shift has brought agriculture from basic methods to highly efficient
practices, thanks to mechanization, innovative technologies, computerized analysis, and
decision-making systems, all of which enhance farming operations and boost crop pro-
ductivity [64]. Innovative machines, commonly referred to as “agribots”, are now widely
utilized in agriculture for various tasks such as soil preparation, seed planting, pest and
weed control, irrigation, fertilization, and even harvesting grains and fruits, significantly
reducing labor and energy costs [71]. Agricultural drones play a key role in overall crop
management, beginning with soil treatment using herbicides, progressing through sowing
and plant treatment with pesticides, conducting physiological monitoring and observation,
and culminating in determining the optimal harvest time [32,72–74]. Today’s agricultural
drones can not only deliver fertilizers, pesticides, water, and herbicides but also capture
images, record videos, and produce real-time maps of fields and crops. This technology
aids farmers in making informed management decisions [75,76].
Farmers now employ drones for livestock monitoring, allowing them to track health
conditions such as illnesses, injuries, and pregnancies with greater precision. The market
for agricultural robots and drones is anticipated to grow significantly, with estimates
projecting it to reach USD 23 billion by 2028 [77]. Geospatial technology, which utilizes
satellite data, enables the application of GIS across various agricultural domains. These
applications encompass crop management, irrigation control, yield forecasting, disease
and weed management, farm automation, livestock monitoring, vegetation mapping, and
predicting land degradation and erosion. GIS is especially effective for precision agriculture,
real-time monitoring, and improving situational awareness, playing a vital role in meeting
the increasing global food demand. Additionally, BC technology is increasingly used to
meet consumer concerns about food origin, quality, and, most importantly, safety [78].

5. Role of AI in Food Industry

AI has been applied across various fields, including modeling, classification, and data
analysis. Example applications include numerical processing, control systems, communica-
tion technologies, interpreting data from pyrolysis, gas chromatography, mass spectrometry,
and HPLC. AI is also used in pattern recognition of RNA, protein, and DNA structures,
predicting microbial growth, biomass, and food shelf life and detecting pathogens [79]. In
the food industry, ANNs, FL, and genetic algorithm techniques have been employed to
enhance performance in handling and engineering processes. AI has been applied in food
science and processing for tasks such as sorting, quality control, and wine analysis. Systems
like Clean-In-Place (CIP) and Clean-Out-of-Place are crucial for maintaining hygiene and
ensuring high product standards in the food industry, often utilizing AI for optimiza-
tion, known as self-optimizing-CIP. However, with the growing global population, food
production remains insufficient due to shrinking agricultural lands, climate change, and
Foods 2025, 14, 411 12 of 38

increasing pollution levels [80]. These factors are negatively impacting food production,
the environment, nutrient availability, and human health. Billions of people worldwide
continue to suffer from nutrient deficiencies. Comprehensive strategies can be employed to
create incentives that drive progress and improve food security outcomes [80].
In response to public demand, farmers have adopted new harvesting techniques that
leverage AI to boost yields [81]. The advancement of horticulture and related technologies
has led to numerous new studies. AI in agriculture involves data analysis, decision-making,
and the application of machine power for the early detection of crop diseases, providing
livestock with optimal nutrition, and enhancing agricultural inputs and profits based on
supply and demand dynamics [79]. Some technologies, such as pest control management
and pesticide data, help farmers increase their yields. A promising solution to these
challenges is the BN. This method is user-friendly and does not require advanced software
skills, making it accessible even to beginners. The Bayesian approach allows for simulations
by integrating prior knowledge into data analysis to produce predictive insights. Unlike
traditional frequentist methods, it does not rely on null hypothesis testing. Instead, it helps
calculate mutual information, which reflects the probability between data sources [80].
By applying these techniques within a BN, the interactions between theoretical factors in
the Global Food Security Index can be analyzed. These Bayesian models are especially
applicable to global food safety, providing intuitive and user-friendly visualizations. AI is
essential in the food industry, contributing not only to quality control and food security but
also to areas like sanitation, manufacturing, and packaging.
As a branch of AI, the CVS merges techniques from image processing and pattern
recognition. This approach is non-destructive, allowing for the analysis and extraction of
features from images, which aids in developing classification patterns [82]. The CVS is
acknowledged as a valuable tool for extracting external feature measurements, including
color, shape, size, and defects. Typically, this system consists of a digital camera, a lighting
setup, and software designed to process images and perform analyses [83]. The system can
be categorized into two types: 2D and 3D versions. Its application spans multiple areas
within the food industry, including predicting color attributes in pork loin [84], assessing
the ripeness of apples [82], determining the roasting level of coffee [85], identifying defects
in pork [86], and evaluating the quality of table grapes [87]. Integrating the CVS with
Foods 2025, 14, x FOR PEER REVIEW 13 of 38
the electronic nose (e-nose) system and soft computing techniques is recognized as a
significant and valuable asset in the food industry (Figure 6). This combination provides
notable benefits, including the ability to achieve accurate predictions rapidly [2]. Table 1
how the integration of the CVS and soft computing has been applied within the food in-
illustrates how the integration of the CVS and soft computing has been applied within the
dustry.
food industry.

Figure
Figure 6. A 6. A schematic
schematic representation
representation system.system.
of e-nose
of e-nose

Table 1. The utilization of the CVS alongside AI in the food industry.

AI Tech-
Application Aims Findings Reference
niques
-The DT-based fuzzy systems can be effectively
utilized for the automated and intelligent clas-
To differentiate be-
-FL sification of Iranian black tea and green tea.
Tea tween Iranian green [88]
-DT -The REP-DT proved to be more advantageous
Foods 2025, 14, 411 13 of 38

Table 1. The utilization of the CVS alongside AI in the food industry.

Application Aims AI Techniques Findings Reference

-The DT-based fuzzy systems can be effectively
utilized for the automated and intelligent
To differentiate between Iranian green tea -FL classification of Iranian black tea and green tea.
Tea [88]
and black tea. -DT -The REP-DT proved to be more advantageous
than the J48 tree for creating a fuzzy
classifier system.
-All the ML algorithms successfully classified
-KNN the beans, with the SVM achieving the highest
To categorize various types of seeds -ANN overall classification rate of 93.13%. This was
Dry beans [89]
based on their production. -SVM followed by the DT, ANN, and KNN, which
-DT recorded classification rates of 92.52%, 91.73%,
and 87.92%, respectively.
-The CVS model, enhanced with various
learning algorithms, utilized the spatial
-SVM
pyramid partition ensemble method for
To forecast the characteristics of barley -DT
Barley flour classifying barley flour. This approach [90]
flour using the improved method. -KNN
achieved accuracies of 75% with a KNN, 95%
-RF
with SVM and an RF, and a perfect score of
100% with the DT method.
-The RBF-ANN model demonstrates superior
classification accuracy compared to FL, with
maximum accuracies of 100% and 88% for the
To automatically assess the ripeness level -FL two models, respectively.
Bell pepper [91]
of bell peppers. -ANN -A system utilizing artificial vision was
developed to predict the maturity of bell
peppers by integrating a CVS with FL and
an ANN.
Foods 2025, 14, 411 14 of 38

Table 1. Cont.

Application Aims AI Techniques Findings Reference

-The CNN integrated with the CVS model
successfully classified apples with an
impressive accuracy of 96.5%. This
For the purpose of differentiating -SVM
Apple performance was significantly better than the [92]
between defective and normal apples. -CNN
traditional image processing method combined
with the SVM classifier, which achieved an
accuracy rate of only 87.1%.
-ANN -The ANN-based model system exhibits a
For the purpose of categorizing bananas -KNN superior classification rate relative to other
Banana [40]
based on their ripeness. -SVM algorithms, achieving an overall recognition
-DT rate of 97.75%.
-The integration of RoboBEER, CVS, and ANN
algorithms facilitated the assessment of beer
production based on customer acceptability
and quality.
To predict beer acceptability based on -ML
Beer -Seventeen ML algorithms were employed to [93]
various sensory parameters. -ANN
identify the best-performing model, with
results indicating that Bayesian regularization
yielded the highest accuracy, achieving an R
value of 0.92.
-The system developed successfully estimated
the weight of Indian sweet lime fruits
with precision.
To estimate the weight of Indian -Various clustering methods were integrated
Lime -ANFIS [94]
lime fruits. with the ANFIS model to enhance accuracy in
the classification system. The results indicated
that FCM was the most effective in predicting
the weight of sweet lime.
Foods 2025, 14, 411 15 of 38

Table 1. Cont.

Application Aims AI Techniques Findings Reference

-The ANN successfully estimated the weight of
the mushrooms, while FL utilized data from the
To assess the quality of mushrooms based -FL
Mushrooms CVS to evaluate the quality of the mushrooms. [95]
on their appearance. -ANN
-The image processing system achieved an
accuracy of 95.6%.
-All models successfully classified the ripeness
-SVM
of cape gooseberries with high accuracy,
To categorize the ripeness of -ANN
Cape gooseberry achieving over 86% accuracy across different [96]
cape gooseberries. -KNN
color spaces. This demonstrates that the system
-DT
is an effective classifier.
-The system developed achieved a success rate
of 97% and an efficiency coefficient of 0.99,
utilizing either two or three input
parameters.-The input parameters for the
Mango To assess the mass of mangoes. -ANN
developed ANN model were derived from the
data obtained from the CVS, enabling the
model to successfully estimate the mass of
the fruit.
-The system’s flexible setup enables
modifications to be made based on the
preferences of individual rice mill operators.
-The automatic control system that was
To manage the operation of rice
Rice -FL developed demonstrated an average [97]
whitening machines.
performance speed that was 31.3% greater than
that of a typical human operator, and it also
enhanced the quality of the output based on
the decisions made by the system.
-The model successfully predicted the quality
To evaluate the quality of pork loin in line
Pork loin -SVM of pork loin by analyzing its color and quality [98]
with industry standards.
attributes in relation to industry demands.
Foods 2025, 14, 411 16 of 38

Table 1. Cont.

Application Aims AI Techniques Findings Reference

-Using the developed system, the ripeness of
To determine the ripeness of
Tomato -ANN the tomato was detected with an accuracy of [99]
fresh tomatoes.
99.31% and a standard deviation of 1.2%.
-The developed system successfully classified
To categorize passion fruits according to
Passion fruits -Multiclases SVM the ripeness levels of passion fruits with an [100]
their ripeness levels.
accuracy of 93.3% in just 0.94128 s.
Key: ANFIS, adaptive neuro-fuzzy inference system; ANN, artificial neural network; CVS, computer vision system; DT, decision tree; FCM, fuzzy C-means; FL, fuzzy logic; KNN,
k-nearest neighbor; ML, machine learning; REP, reduced error pruning; RF, random forest; SVM, support vector machine.
Foods 2025, 14, 411 17 of 38

5.1. AI in 3D Printing
3D printing refers to a collection of additive manufacturing processes that originated
in the late 1980s, transitioning from prototyping tools to a disruptive technology ecosystem.
While the methods may vary, they share common applications, primarily involving the
construction of material layers. For instance, metal powder is melted using lasers, or
liquid polymers are solidified with ultraviolet light to create the desired structure. This
technology enables the creation of complex shapes and allows for the generation of digital
models, facilitating decentralized production and customization of parts [101]. 3D printers
function similarly to robots. The increase in robotics manufacturing is driven by the ease of
electronics development, the accessibility of powerful cloud computing, and the availability
of high-quality sensors. In the past, robots were prohibitively expensive and primarily
utilized in large-scale industries such as automotive manufacturing, where companies
would typically invest at least USD 1,000,000 to incorporate them into production lines [102].
3D printing and robotics are significantly transforming the confectionery industry, offering
faster and more efficient operations [79].
The process begins with a digital file created in a computer-aided design software
environment, where every aspect of the model can be precisely defined, modified, and
optimized using 3D design tools. These advancements have captured the attention of com-
panies seeking streamlined production methods [103]. Once the digital model is complete,
it is saved as a .stl file and then transferred to the 3D printing software interface. Many 3D
printing programs are open source and free to use. The software slices the object into layers
and generates individual commands. Key parameters such as speed, temperature, height,
and thickness, often controlled by robotic arms, are configured within the 3D printing
software. The completed design and its features are then exported in a format compatible
with the 3D printer for production [104].
A state-of-the-art 3D production system, equipped with a culinary innovation center
utilizing 3D printing technology, has been introduced to allow chefs in the food industry
to experiment with the ChefJetTM Pro professional food printer (3D systems, California,
USA). In partnership with Hershey, a prototype 3D chocolate printer called CocoaJet
(Cocoapress, Florida, USA) was unveiled in 2015. In Australia, TM Retail Food Group
began implementing 3D chocolate printing for personalized cake messages within their
Michel’s Patisserie franchise, with plans to roll it out nationwide by 2018 [105,106]. After
making modifications to commercial printers, the equipment successfully obtained a “food
grade” certification from the Federal Agency for the Safety of the Food Chain.
The use of software is a critical factor in designing, slicing, and optimizing the model.
Slicing software (ver 1.0) serves as a bridge, enabling the planning and assessment of
the layers between the 3D model and the 3D printer. This software converts the digital
model into a physical form by translating .stl files into g-code for 3D printing. Setting
up the software is relatively simple, requiring only a few adjustments. Key factors such
as nozzle temperature, printing speed, layer thickness, and platform temperature are
essential to optimize the printing process, along with features for support design and
model repair [107].

5.2. AI in the Dairy Sector

Milk is transported from various production centers to processing plants via milk
trucks. Once at the plant, the milk undergoes pasteurization, a process that eliminates
bacteria and extends the milk’s shelf life. The pasteurized milk is then used to produce a
variety of dairy products. For optimal quality, milk should have a pH value of 6.7. If the
pH deviates from this level, the milk spoils rapidly. Refrigeration helps prevent spoilage,
and this process is typically carried out at milk stations to preserve the product until
Foods 2025, 14, 411 18 of 38

it reaches the collection point. To address this issue more effectively, a wireless sensor
network integrated with AI technology is needed to monitor and maintain the quality of
the milk throughout the supply chain [108]. AI also plays a significant role in automating
the milking process in the dairy industry. Traditionally, manual milking is time-consuming,
but this can be minimized with the use of Automated Milking Systems or milking robots.
Although this technology was introduced in the US in 2000, it was first developed in Europe
in 1992 [109]. By 2002, nearly 1754 milking robots were in use, a number that grew to 8190
within five years, and by 2010, the total had risen to 16,000 [109].
In 2010, Germany and France accounted for 30% of the robots in operation [110].
According to experts, the robotic milking market is expected to support 28,600 robots
annually. Cows are trained for the automated milking process and equipped with electronic
tags, enabling the robot to recognize each cow and dispense feed. The robot attaches
milking cups to the teats, initiating the milking procedure. As each quarter completes
milking, the cups are automatically removed, and a disinfectant is applied before the cow
leaves the milking station [109]. Robots also conduct disease detection tests on milk using
technologies such as laser scanners, ultrasound, and the Optical Guidance System. When
no signs of disease are detected, the milk is directed for cooling [110]. These robots utilize
a CVS for navigation and sensing, while ML, a critical aspect of AI, allows them to learn
from human interactions [109]. Programmed CIP systems handle the cleaning process with
two distinct programs: (a) a CIP program for rotary cleaning, which includes equipment
such as heating surfaces and pasteurizers, and (b) a CIP program for rotational cleaning,
which includes tanks used for storing purified milk [79].
With AI integration, operators can select specific steps and run programs based on
signals from the system, allowing them to monitor temperature levels for different liquids in
various tanks. Unified CIP systems are commonly used in large milk sectors for separation
processes. In the dairy industry, robots are employed for cheese packaging, cutting, and
curd slicing according to customer-specified shapes and sizes. Special grippers enable the
robots to pick up cheese blocks and place them on conveyors for further processing [109].

5.3. AI in Food Processing Strategies

The vast amount of data surrounding food and edible products have driven researchers
to investigate this field through the application of AI. By 2015, computers had advanced
to the point where they could identify food from images. By early 2016, MIT’s AI had the
capability to predict ingredients and assess nutritional values based on the food presented
to it. This technology quickly reached the public, becoming available as a mobile application
within a few months [111]. AI technologies assist the food industry in efficiently promoting
products to the market through global food trend strategies and planning [112]. Food
processing becomes more adaptable with machines capable of handling tasks ranging from
simple distinctions, like identifying apples versus oranges, to more complex tasks, such
as differentiating between low-saturated and high-unsaturated fats. Figure 7A illustrates
food monitoring techniques used at “Stemmer Imaging (Puchheim, Germany” for various
applications, while Figure 7B demonstrates the application of DL in food classification
using the LeNet architecture.
 fats. Figure 7A illustrates food monitoring techniques used at “Stemmer Imaging
(Puchheim, Germany” for various applications, while Figure 7B demonstrates the appli-
Foods 2025, 14, 411 cation of DL in food classification using the LeNet architecture. 19 of 38

Figure 7.7.AIAIand
Figure andcomputer
computervision
vision applications in the
applications in thefood
foodindustry:
industry:(A)
(A) harvesting,
harvesting, quality
quality control,
control,
picking, and sorting utilizing vision algorithms; (B) food classification with LeNet-DL architecture.
picking, and sorting utilizing vision algorithms; (B) food classification with LeNet-DL architecture.

5.4. AI in Food Safety

5.4. AI in Food Safety
The sterile characteristics of robots make them highly suitable for use in food pro-
The sterile characteristics of robots make them highly suitable for use in food pro-
cessing industries, significantly contributing to the reduction in foodborne diseases. This
cessing industries, significantly contributing to the reduction in foodborne diseases. This
aspect
aspectisisparticularly
particularlyimportant
important in in light of the
light of the stricter
stricterhygiene
hygienerequirements
requirements outlined
outlined in in
the Food Safety Modernization Act, which applies to the entire supply chain.
the Food Safety Modernization Act, which applies to the entire supply chain. The primary The primary
concern
concernisisthat
thatproducts
productslike
like cereals,
cereals, spices, and other
spices, and othershelf-stable
shelf-stablefoods,
foods,which
whichdodo notnot
require refrigeration, are particularly vulnerable to contamination. While
require refrigeration, are particularly vulnerable to contamination. While these foodsthese foods were
once
wereconsidered less prone
once considered to contamination,
less prone to contamination,thethe
situation has
situation dramatically
has dramaticallychanged
changed in
recent times.
in recent AI-based
times. AI-based systems
systemsoffer
offeraapromising
promising solution
solution totothese
thesechallenges,
challenges, as as they
they
are not susceptible to transmitting illnesses in the same way humans can.
are not susceptible to transmitting illnesses in the same way humans can. Furthermore, Furthermore,
maintaining
maintainingAI-driven
AI-driven systems
systems is straightforward
straightforward and andefficient,
efficient,making
makingthem
them anan ideal
ideal
choice for ensuring hygiene and reducing the risk of contamination in food processing
environments [113]. A report by Technavio indicates that the adoption of robots in food
processing industries is expected to increase by 30%, aligning with government regulations.
Additionally, several innovative applications of AI in food safety are emerging and are
expected to gain widespread recognition in the near future. These advancements aim to
significantly reduce the occurrence of foodborne diseases. Figure 8 illustrates the role of AI
in enhancing food safety [114].
 tions. Additionally, several innovative applications of AI in food safety are emerging and
are expected to gain widespread recognition in the near future. These advancements aim
Foods 2025, 14, 411 to significantly reduce the occurrence of foodborne diseases. Figure 8 illustrates the role
20 of 38
of AI in enhancing food safety [114].

Figure
Figure 8. AI in
8. AI in food
food safety.
safety.

5.4.1. Next-Generation Sequencing and Electric Noses

5.4.1. Next-Generation Sequencing and Electric Noses
Two of the most promising innovations in the food industry are next-generation se-
Two of the most promising innovations in the food industry are next-generation se-
quencing (NGS) and ENs. NGS is rapidly replacing traditional DNA-based methods in
quencing (NGS) and ENs. NGS is rapidly replacing traditional DNA-based methods in
food safety applications. The integration of AI-powered automated systems and workflows
food safety applications. The integration of AI-powered automated systems and work-
has significantly accelerated data collection and laboratory testing, making these processes
flows has significantly accelerated data collection and laboratory testing, making these
faster and more accurate than ever before. NGS is highly effective at quickly detecting
processes faster and more accurate than ever before. NGS is highly effective at quickly
potential hazards, enabling the prevention of widespread infections before they impact a
detecting potential hazards, enabling the prevention of widespread infections before they
large number of individuals. ENs function as substitutes for human olfactory senses in
impact a large number of individuals. ENs function as substitutes for human olfactory
manufacturing environments. These devices are equipped with sensors that can accurately
senses ainwide
detect manufacturing environments.
range of odors. The sensorsThese devices
capture theare equipped with
surrounding sensors
smells, andthatthe can
col-
accurately detect a wide range of odors. The sensors capture the surrounding
lected data are transmitted to a centralized data center, where ML algorithms analyze the smells, and
the collected [113,115].
information data are transmitted
Based on the to aanalysis
centralized data
of the center, where
ML-based system, MLanalgorithms
alert is sent ana-
to
lyze the information [113,115]. Based on the analysis of the ML-based
the production units when necessary. As a result, ENs have the potential to become a key system, an alert is
sent to the production units when
technology in the future of food safety. necessary. As a result, ENs have the potential to become
a key technology in the future of food safety.
5.4.2. Food Waste Management
5.4.2. Food Waste Management
According to a report from the U.S. Department of Agriculture, food waste in the
UnitedAccording to a reporttofrom
States is estimated the for
account U.S.30Department
to 40% of the oftotal
Agriculture,
food supply.foodThis
waste in the
estimate,
United States
derived from isthe estimated
USDA’stoEconomic
account for 30 to 40%
Research of the indicates
Service, total food that
supply.31%This estimate,
of food loss
derived from the USDA’s Economic Research Service, indicates
occurs at the retail and consumer levels, which equates to approximately 133 billionthat 31% of food loss oc-
curs at the
pounds andretail
USD and161consumer levels,ofwhich
billion worth food equates
in 2010.toThis
approximately
level of waste 133 has
billion pounds
significant
and USD 161 billion worth of food in 2010. This level of waste has
widespread social implications [114]. Moreover, AI has the potential to address significant and wide-
spread
food social
waste implications
issues and create [114]. Moreover,
significant AI has theby
opportunities potential
reducing towaste
address food
levels bywaste
2030.
issues and create significant opportunities by reducing waste
These impressive outcomes can be achieved through the implementation of additionallevels by 2030. These im-
pressive outcomes
regenerative can bepractices
agricultural achieved[116].
through the implementation of additional regenera-
tive agricultural practices [116].
This indicates that resources are not being utilized optimally. Traditional farming
Thiscan
methods indicates that resources
be replaced are not being
by more advanced, utilized
intelligent optimally.
farming Traditional
techniques. farming
These methods
methods can be replaced by more advanced, intelligent farming
involve the use of various sensors to gather data which are then processed using techniques. These meth-
ML
ods involvetothe
algorithms use informed
make of variousdecisions.
sensors toBy gather data these
adopting whichapproaches,
are then processed
farmers using
can makeML
algorithms to make informed decisions. By adopting these approaches,
faster and more accurate decisions. Below are some suggested strategies to reduce food farmers can make
faster through
waste and more theaccurate
use of AI:decisions. Below are some suggested strategies to reduce food
waste
i. through the use of AI:
While some studies focus on assessing the ripeness of fruits, others highlight the
i. While
role some studiesmicroorganisms
of beneficial focus on assessing the ripeness
in promoting ofgrowth
the fruits, others highlight
of fruits the role
and vegetables
of without
beneficial microorganisms
the in promoting
need for artificial fertilizers; the growth of fruits and vegetables with-
ii. outProducers
the need for
canartificial
eliminatefertilizers;
the need for soil testing by leveraging the benefits of AI,
ii. Producers can
leading to eliminatecost
significant thesavings;
need for soil testing by leveraging the benefits of AI,
iii. leading to significant
Farm-based cost savings;
food supply chain management employs CVSs to monitor and analyze
each step of the process, leading to a significant reduction in food waste;
iv. AI-driven food tracking systems will allow for the sale of food before it becomes
waste. This approach facilitates greater connectivity between farmers and con-
sumers, promoting more efficient food distribution.
The implementation of such concepts cannot be achieved by a single organization or
entity alone. Transforming the entire food industry is essential, requiring a collaborative
effort from a network of stakeholders. A unified approach is necessary to develop an
efficient system that can have a significant global impact.
Foods 2025, 14, 411 21 of 38

5.5. AI in Beverage and Soft Drinks

Drinks can be divided into three main types: (i) alcoholic beverages, (ii) non-alcoholic
beverages, and (iii) hot beverages. Alcoholic drinks consist of beer, wine, and spirits,
while non-alcoholic options include milk, soda water, juices, and soft drinks like Thumbs
Up, Miranda, and Coca-Cola. Hot beverages cover items such as coffee, tea, and hot
chocolate [117]. The process of beer fermentation starts by introducing yeast to the aerated
wort. The yeast feeds on the nutrients in the wort, initiating its growth and simultaneously
generating alcohol and other metabolites. Fermentation proceeds until the sugar levels
in the wort drop to a specified concentration, leading to increased alcohol production
as the wort becomes lighter. Crucial factors that must be controlled during this process
are oxygen levels, temperature, and the yeast pitching rate. Other factors that impact
fermentation include wort composition and yeast condition. Inconsistencies in yeast and
wort can be mitigated by adjusting the pitching rate or slightly increasing oxygen levels and
temperature when yeast viability is reduced. In traditional brewing, continuous monitoring
is challenging, but with the use of AI and its programmed tools, the brewing process can
now be effectively monitored and controlled in real-time [118].

5.5.1. AI in Beer and Wine Processing

AI approaches such as CVSs, FL, neural networks, and hybrid intelligence methods
are used to control the beer brewing process. Beer quality can be predicted by applying
FL rules to real process data, which helps in detecting higher alcohols, vicinal diketones,
and fatty acids. Additionally, ANNs assist in monitoring and reporting the progress
of beer fermentation. Beer is among the most ancient and widely consumed alcoholic
drinks, appreciated by one-third of the global population. It possesses two main types
of characteristics: sensory and visual. The visual aspects include color, clarity, turbidity,
foam volume and retention, and overall appearance. On the sensory side, attributes such
as aroma, bitterness, and mouthfeel are key [35]. Traditionally, evaluating these criteria
requires significant manpower and time. However, using the CVS approach, this process
can be automated. CVS inspects the beer’s visual qualities through digital photographs,
utilizing components such as a camera, lighting, and a computer. It performs tasks like
image capture, processing, and analysis to detect objects and extract qualitative data from
samples. A crucial next step is digitization, where the image is converted into numerical
data for further evaluation [35].
AI-powered mobile software for CV and image analysis is utilized to detect grape
clusters that are unsuitable for wine production due to spoilage. This process is carried
out in three essential steps. The initial step is image acquisition, where photographs are
captured using mobile software in the vineyards one week prior to harvest. This allows
for the evaluation of the grape clusters’ compactness. Expert panels can then manually
rate the compactness based on guidelines provided by the International Organization of
Vine and Wine (OIV 204) [119]. Cluster compactness is rated across nine categories, as
shown in Figure 9. The second step is image processing, which serves as a precursor to the
third step, segmentation. This segmentation step requires detailed data for each cluster to
identify rotten or substandard clusters. Various methods and algorithms, such as K-means,
Gaussian, and cross-validation, are employed to analyze and classify the clusters based on
their quality [120].
 the third step, segmentation. This segmentation step requires detailed data for each cluster
to identify rotten or substandard clusters. Various methods and algorithms, such as K-
Foods 2025, 14, 411 means, Gaussian, and cross-validation, are employed to analyze and classify the clusters
22 of 38
based on their quality [120].

Figure 9. Grape
Figure 9. Grape cluster
cluster classification.
classification.

5.5.2. ANN in Non-Alcoholic Beverage

5.5.2. ANN in Non-Alcoholic Beverage
A specialized type of ANN, the Deep-CNN, is employed for the nutritional analysis
A specialized type of ANN, the Deep-CNN, is employed for the nutritional analysis
of soft drinks to assist in managing weight gain and obesity. This approach estimates the
of soft drinks to assist in managing weight gain and obesity. This approach estimates the
nutritional content by evaluating factors such as bottle size and the ratio of the cap. This
nutritional content by evaluating factors such as bottle size and the ratio of the cap. This
method is applicable to both carbonated and fruit-based beverages. By incorporating image
method is applicable to both carbonated and fruit-based beverages. By incorporating im-
processing within the CNN framework, the background is eliminated from the image,
age processing within the CNN framework, the background is eliminated from the image,
enabling precise calculation of the nutritional content [121].
enabling precise calculation of the nutritional content [121].
5.5.3. ANN in Hot Beverage
5.5.3. ANN in Hot Beverage
The EN is a device developed to detect and evaluate odors, working similarly to the
The ENolfactory
mammalian is a device developed
system to detect
by utilizing and
gas evaluate
sensors odors, It
[122,123]. working similarly
is extensively to the
used in
mammalian olfactory system by utilizing gas sensors [122,123].
industries like wine and coffee for detecting scents. In the food and beverage industry, It is extensively used in
it helps monitor and ensure consistent product quality. The quality of roasted coffee isit
industries like wine and coffee for detecting scents. In the food and beverage industry,
helps monitor
assessed through and theensure
coffee consistent productinquality.
cupping method, accordanceThe with
qualitytheof roasted coffee
standards is as-
established
sessed through the coffee cupping method, in accordance with the
by the Technical Standards Committee of the Specialty Coffee Association of America [122]. standards established
by the
First, 5 gTechnical
of coffee Standards Committee
beans are ground into aoffine
thepowder
Specialty andCoffee
steeped Association
in boiling of America
water for 3
[122]. First, 5 g of coffee beans are ground ◦ into a fine powder and
to 5 min at a temperature range of 93–97 C [124]. The cupping method is used to collect steeped in boiling water
for 3 which
data, to 5 min areatthen
a temperature rangeaofradial
predicted using 93–97basis
°C [124]. The ANN.
function cupping method
Signals from is the
usedgas to
collect data, which are then predicted using a radial basis function
sensors in the EN are analyzed by a computer through LabVIEW software, with the data ANN. Signals from the
gas sensors
further in theusing
processed EN are analyzed byanalysis
multivariable a computer through
techniques. The LabVIEW
experiment software, withinto
is divided the
data further processed using multivariable analysis techniques. The
three phases: first, examining the effect of temperature (ranging from 10 C to 90 C) on experiment
◦ is divided
◦
intosmell
the threecharacteristics
phases: first, examining the effect
of liquid coffee; of temperature
second, categorizing (ranging
the acidityfromlevels
10 °C of to coffee
90 °C)
on the smell characteristics of liquid coffee; second, categorizing the
based on various degrees of roasting; and finally, testing bitterness levels through human acidity levels of coffee
based on various
evaluation, guided degrees of roasting;
by a radial and finally,
basis function ANN testing
[122].bitterness
Like thelevels through
e-nose, human
the e-tongue
evaluation, guided by a radial basis function ANN [122]. Like the
(Figure 10) is also used to assess the quality of various beverages, including milk, coffee, e-nose, the e-tongue
(Figure
tea, wine,10)and
is also
beer.used
The to assess the
e-tongue quality
detects keyof various beverages,
parameters including
such as saltiness, milk, coffee,
sourness, and
tea, wine, and beer. The e-tongue detects key parameters such as
bitterness [125]. The quality of tea can be evaluated using an e-tongue, which measures saltiness, sourness, and
key flavor components such as theaflavin and thearubigin. Pu-erh tea samples of various
grades and origins have been analyzed using the e-tongue along with other instrumental
methods, including chemical analysis and electronic tongue technology [126]. A pulse
 grades and origins have been analyzed using the e-tongue along with other instrumenta
methods,[125].
bitterness including chemical
The quality analysis
of tea can be and electronic
evaluated usingtongue technology
an e-tongue, which[126]. A pulse vol
measures
tametric
key flavor e-tongue
componentsis such
usedastotheaflavin
determineandthe type of tea
thearubigin. based
Pu-erh on its age,
tea samples which can be
of various
Foods 2025, 14, 411 23 of [127].
38
achieved through analysis using UV-VIS spectrophotometer-based equipment
grades and origins have been analyzed using the e-tongue along with other instrumental
methods, including chemical analysis and electronic tongue technology [126]. A pulse vol-
tametric e-tongue is used to determine the type of tea based on its age, which can be
voltametric e-tongue is used to determine the type of tea based on its age, which can be
achieved through analysis using UV-VIS spectrophotometer-based equipment [127].
achieved through analysis using UV-VIS spectrophotometer-based equipment [127].

Figure 10. A schematic representation of e-tongue system.

Figure
Figure 10.
10. A
A schematic
schematic representation
representation of e-tongue system.
The classification of coffee beans follows specific standards, taking into account cat
egory,Thedefects,
The quality
classification characteristics,
coffee beans follows
of coffee andspecific
the nature of thetaking
standards, resulting
taking intobeverage
into account [128]. Ara
account cate-
cat-
bic espresso
gory,
egory, defects, types
defects,quality are categorized
qualitycharacteristics,
characteristics, and based
andthe on type
thenature
nature ofoftheorresulting
the defects,beverage
resulting ranging[128].
beverage from two to eigh
Arabic
[128]. Ara-
classifications.
espresso
bic types
espresso are
typesWhen assessing
categorized
are categorizedbasedquality, onitor
on type
based isdefects,
type important ranging
or defects, to consider
from
ranging two
from factors
to that
eightinfluence
eighttoclassifi-
two
cations. When
market perception
classifications. assessing
Whenand quality,
enhance
assessing it is important
the final
quality, to consider
product’sto
it is important factors
value. that
The factors
consider influence
classification market
process is par
that influence
perception
market and
perception enhance
and the
enhance final
the product’s
final value.
product’s The
value. classification
The process
classification
of the business evaluation, focusing on aspects such as shape, size, color, grain is
process part of
is part
types, and
the
of business
the business
beverage evaluation, focusing
evaluation, focusing
characteristics, on aspects
on aspects
all of which such as shape,
such as shape,
are integral size,
to thesize, color, grain types,
color, cycle (Figure
marketing and
and 11). By
beverage characteristics,
beverage characteristics, all
all of
of which
which areare integral
integral toto the marketing
marketing cycle (Figure 11). By By
utilizing image analysis, defects in the beans can be identified based on their size and
utilizing image analysis, defects in the beans can be identified based
utilizing image analysis, defects in the beans can be identified based on their size and on their size and
shape, allowing for the removal of defective beans and ensuring the product’s quality is
shape, allowing
shape, allowing for
for the
the removal
removal of of defective
defective beans
beans and
and ensuring
ensuring thethe product’s
product’s quality
quality is is
maintained [128].
maintained [128].
maintained [128].

Figure 11. Defect detection using image processing algorithms.

Figure 11.Defect
Figure11. Defectdetection using
detection image
using processing
image algorithms.
processing algorithms.
Foods 2025, 14, 411 24 of 38

5.6. AI in Bakery
Automation is advancing rapidly across industries, including the baking sector [129].
Bakery products come in a wide variety of shapes, sizes, and forms, making grading a
crucial part of the production process, especially during handling and packaging where
irregularities often occur [123]. In the past, quality assurance relied on human visual
inspection. However, in modern bakery production, bread must undergo several processes
to ensure consistency and quality. Bread making typically involves the introduction of live
microorganisms, such as yeast, into the dough. If any step in the bread-making process
is not executed correctly, the quality of the final product will be compromised [79]. In the
bakery industry, a key focus is on optimizing production processes, utilizing resources
efficiently, and implementing automatic control systems (ACSs) to improve product quality,
lower costs, and boost profitability. The implementation of ACSs in bread-making enhances
both productivity and efficiency, while also lowering electricity and fuel consumption
during essential production processes. Additionally, optimizing resource efficiency and
ensuring accurate information in bread production are crucial approaches for further
improving operations [130].
Robotic technology systems in baking include automated control systems that monitor
and regulate various critical factors. These sensors regulate critical parameters such as
the rheological properties of dough and sourdough, the rising potential of the dough,
the active acidity of the sourdough, the dough’s acidity and aroma, the dough’s shaping
capabilities, the proofing time and temperature, humidity in the proofing chamber, and the
dough’s weight [130]. The bread’s acidity, porosity, moisture content, dimensional stability,
and core temperature, as well as the baking time, are implicitly controlled by a system
of sensors along with a visualization interface. An intelligent decision-support system
processes sensor data using components such as databases, knowledge bases, training
modules, output blocks, and ESs, allowing for adjustments to the operational modes of
ultrasonic systems for optimal bread production. This is accomplished by employing
control mechanisms that enhance the management of effects across various technological
environments. The ES for product quality control evaluates raw material parameters,
enhancing the properties of dough, flour, and sourdough to improve the fortification
qualities of the bread [79].
The entire baking process, from mixing to packaging, typically takes about three hours.
Product quality assessment usually takes an additional hour, meaning the quality data
cannot be immediately fed back to the mixer, oven, or other devices. Sensors designed
for food quality play a crucial role by converting responses related to food properties into
electrical signals, facilitating monitoring and control [131]. Sensors can operate in either
online or offline modes. To assess bread quality, the inner portion of the loaf is typically ex-
amined using a camera after cutting. The quality of the loaf is assessed by measuring factors
like dough porosity and size. These devices help monitor bakery products by delivering
real-time data analysis according to the volume and variety of bread produced [79].

5.7. AI in Drying Fresh Foods

Advanced drying techniques that use physical fields, including microwave drying, ra-
dio frequency (RFQ) drying, infrared radiation (IR) drying, and ultrasonic drying, provide
notable benefits for processing fruits and vegetables. These techniques can substantially
reduce energy consumption and drying time, while also enhancing the drying process.
Additionally, these methods help maintain the sensory attributes and nutritional value
of fruits and vegetables, enhancing the overall quality of the dried products. Although
physical field drying offers several benefits, it also poses challenges, including uneven
drying, aroma loss, and nutrient degradation. However, these limitations can be miti-
Foods 2025, 14, 411 25 of 38

gated by integrating AI into the drying process. AI-assisted physical field drying offers a
promising solution for improving the drying of vegetables and fruits. Among the various
physical field techniques, microwave drying stands out as a highly efficient and commonly
applied method. By leveraging AI, the performance and outcomes of microwave drying
can be further optimized. During the food drying and heat transfer process, the technique
allows heat to penetrate materials and products without the need for an additional ther-
mal gradient. This enables more efficient heat distribution throughout the product being
dried [132]. Microwave technology has the potential to greatly improve both the drying
rate and efficiency in the processing of vegetables and fruits. Research on potato chips has
shown that microwave drying, particularly when combined with treatments using sodium
chloride and sucrose, can substantially reduce the overall drying time [133]. Monton
et al. [134] found that integrating convection drying with microwave drying significantly
shortened the drying time of turmeric, while still meeting the quality standards set by the
Thai Pharmacopoeia. Following the drying process, the moisture content was lowered
to below 10.0% v/w, the volatile oil content surpassed 6.0% v/w, and curcumin levels
exceeded 5.0% w/w. This combination of drying methods resulted in higher-quality dried
products compared to using a single drying technique. Despite its advantages, microwave
heating has certain drawbacks, such as uneven heating, shallow penetration depth, and the
occurrence of phenomena like “expansion” [135]. To address these issues, Lv et al. [136]
introduced an advanced intelligent microwave vacuum drying (MVD) system, incorpo-
rating low-field nuclear magnetic resonance technology to address these challenges. This
device enables real-time monitoring of water content during the drying process of fruits
and vegetables, offering more precise control and optimization. This system establishes a
relationship between the water content of fruits and vegetables (M2) and the amplitude of
the fitting signal (A2). The development of a linear model connecting M2 and A2 allows
for accurate estimation of the drying endpoint in MVD, achieving a precision level greater
than 95% (p > 0.950). It effectively monitors changes in water content during the microwave
drying process. Additionally, Chen et al. [137] applied microwave drying to honeysuckle
and introduced an FL control method for parameter regulation. This approach efficiently
controlled the drying temperature, resulting in optimal drying outcomes. During the
drying process, moisture content can be monitored using intelligent sensor technology,
while the drying temperature can be regulated through the application of FL. This combi-
nation helps address the limitations of microwave drying, allowing for the determination
of optimal drying parameters and improving product quality. Additionally, RFQ drying, a
volumetric heating method, enables rapid and deep penetration of heat into food, further
enhancing the drying process [138]. The capability of RFQ drying to penetrate food more
thoroughly than microwave drying is attributed to its free-space wavelength, which is 20
to 360 times longer than that of microwaves. RFQ drying not only reduces drying time but
also improves overall efficiency. To enhance drying speed, efficiency, and product quality,
Zhou et al. [139] conducted a study using an RFQ vacuum drying system on kiwifruit
slices, operating at a frequency of 27.12 MHz and a power level of 3 kW. The dried kiwifruit
displayed vibrant color, high vitamin C content, and remarkable rehydration properties.
CVSs are effective tools for monitoring changes in shrinkage and color rate during the
drying process of vegetables and fruits. By integrating a CVS into the RFQ drying process,
issues related to color, nutritional content, and sensory quality can be effectively managed.
Additionally, the application of control and analysis technologies, such as ANNs and FL,
allows for the precise regulation of time and power, leading to improved drying outcomes.
IR radiation, characterized by its spectrum and directionality, is a type of electromag-
netic radiation that falls within the wavelength range of 0.78–1000 µm [140]. IR drying
offers several advantages, including rapid transient response, compact equipment, effi-
Foods 2025, 14, 411 26 of 38

cient convection and conduction, and significant energy savings, compared to microwave
heating. As a result, it has become a key heat treatment technology in food processing,
providing an effective and energy-efficient method for drying [140]. IR drying is beneficial
for vegetable and fruit drying as it enhances the drying rate while helping to preserve their
nutritional content. Adak et al. [141] investigated the influence of air temperature, drying
conditions, and infrared power on the drying of strawberries. Their findings revealed that
increasing infrared power reduced the drying time, while also improving the nutritional
value of the strawberries, including higher antioxidant capacity and better retention of total
anthocyanins and phenolic compounds. Although IR drying offers several advantages,
its use in isolation can occasionally harm the sensory qualities of fruits and vegetables.
Mihindukulasuriya et al. [142] observed that infrared drying of red pepper resulted in
significant weight loss, a reduction in redness, and a decrease in capsaicin content. Such
undesirable changes are not typically expected in the drying of vegetables and fruits. How-
ever, incorporating AI-assisted control within the drying process allows for more effective
management of changes in color and volume, thus addressing these challenges. An intelli-
gent fuzzy machine vision control system (FMCS) was developed by Nadian et al. [143],
which combines FL and a CVS to regulate operational variables during infrared drying
using mixed hot air. This system monitors the color and volume changes during kiwifruit
drying and uses a genetic algorithm to optimize the quality of the dried products. The
FMCS significantly reduces drying time compared to conventional hot air drying, cutting
it from 40 min to 24 min (a 40% reduction). Additionally, it enhances product quality
compared to standalone infrared drying, as evidenced by a reduction in color change from
7.9 to 2.1, a more than threefold improvement. The advancement of novel pretreatment
and post-treatment technologies offers potential for significantly improving the quality of
dried vegetables and fruits [144].
Ultrasound transmission through a medium generates various physical and chemical
effects, making it a valuable tool in food drying applications [145]. Ultrasound, when
applied to vegetables and fruits, can aid in preserving product quality after the drying
process. Marcela et al. [146] found that using either an ultrasonic probe (direct method)
or an ultrasonic bath (indirect method) for pretreating beet snacks significantly decreased
color changes and cyanogen content, while also cutting the total processing time by ap-
proximately 26%. Wang et al. [147] investigated the effects of ultrasonic pretreatment on
carrots before intermediate-wave IR drying. Their findings revealed that carrots subjected
to ultrasound prior to IR drying preserved higher levels of β-carotene and demonstrated
a superior rehydration ratio when compared to the untreated control samples. While
ultrasonic treatment can enhance certain drying characteristics, it may also negatively
affect the texture of vegetables and fruits. In their research, Rodriguez et al. [148] reported
significant alterations in the microstructure of apples resulting from ultrasonic treatment
during the drying process. However, by integrating AI technology to regulate ultrasonic
power and control the rate of water loss during drying, these issues can be minimized.
Additionally, Pu et al. [149] employed near-IR imaging to monitor the moisture distribution
in mango slices during drying, which facilitated more uniform moisture removal and
helped preserve the shape of the mango slices. In addition to addressing the limitations
of various physical field drying methods for fruits and vegetables, AI technology can
assist in optimizing the drying process. It provides valuable insights into the dynamics
of physical field drying, enabling the development of more precise and accurate drying
models. Taghinezhad et al. [150] conducted a study on the microwave drying of papaya,
utilizing an ANN, particle swarm optimization, and the gray wolf optimization (GWO)
algorithm to predict the relevant drying model. The GWO algorithm is a novel swarm
intelligence optimization technique that emulates the leadership and hunting behaviors
Foods 2025, 14, 411 27 of 38

observed in gray wolves in their natural environment. This approach demonstrated strong
analytical capabilities, accurately predicting the drying model of papaya. The results
showed that this optimization method yielded the highest R2 value (0.9707) for the effective
diffusion coefficient. Dai et al. [151] developed a combined IR convection (IRC) dryer that
utilizes a support vector regression (SVR) algorithm, enhanced by an improved particle
swarm optimization (IPSO) method. This system is capable of modeling and controlling the
nonlinear behavior of drying particles. They proposed an IPSO-SVR model to manage and
predict the drying process of grains using IRC, providing more accurate control over the
drying dynamics. The development of these models enables better control and monitoring
of the drying process, allowing for precise regulation of drying parameters. This not only
helps reduce costs but also ensures the production of high-quality dried products. Table 2
provides an overview of other AI technologies applied to assist in the physical field drying
of fruits and vegetables.

Table 2. Advanced drying technology utilizing high-efficiency physical fields for AI applications.

Products AI Technology Drying Technology Drying Condition Reference

-Temperature: 40, 50, and 60 ◦C
-Pressure: 0.8 bar
Moringa olifera leaves ANN Vacuum tray dryer [152]
-Dryer plates: 0.44 × 0.44 m
Time: 30 min
-Temperature: 40 ◦ C
Watermelon rind -Energy consumption:
ANN Solar air heaters [153]
pomace 609 kWh.kg−1 and
318 kg CO2 .kWh−1
Stevia rebaudiana Leaves ANN Microwave drying -Power: 180 to 900 W [154]
-Temperature: 50–80 ◦C
Green tea leaves ANN Fluidized bed drying [155]
-Air flow velocity: 7–9.5 m/s
-Temperature: 55, 60, 65, 70, and
Hot air impingement 75 ◦ C
Mushroom slices ANN [156]
dryer -Air velocity: 3, 6, 9, and 12 m/s
-Sample thickness: 6, 9, and 12 mm
-Temperature: 40 ◦ C, 45 ◦ C, 50 ◦ C,
and 55 ◦ C
Air-impingement
Walnut ANN -Air velocities: 1, 2, 3, and 4 m/s [157]
technology
-Moisture content: 10, 15, 20,
and 25%
-Air velocity: 3.58, 4.25, and
Conventional
4.82 m/s
Potato slices ANN multi-stage convective [158]
-Temperature: 45 and 50 ◦ C
cabinet dryer
-Time: 240 min
-Temperature: 20 ◦ C
-Time: 28.54 min
-Osmotic concentration: 55.58%
Osmo-sonicated
Garlic slices ANN -Optimum responses: WL = 25.839%,
dehydration
SG = 3.557%, RR = 7.512,
DR = 0.163 g H2 O/g.dm/min, and
AC = 15.229 mg/g
-Temperature: 50, 60, and 70 ◦ C
Convective and
Apple ANN -Air velocity: 1.0 m/s [159]
microwave
-Power: 90, 180, and 360 W
Foods 2025, 14, 411 28 of 38

Table 2. Cont.

Products AI Technology Drying Technology Drying Condition Reference

Microwave–hot air -Temperature: 23, 50, and 70 ◦ C
Mushroom ANN [160]
dryer -Power density: 1.5, 2, and 2.5 W/g
Ginkgo biloba
ANN Microwave drying -Power: 200, 280, 460, and 640 W [161]
seeds
Custom designed -Temperature: 40, 50, and 60 ◦ C
Onion ANN [162]
fluidized bed dryer -Air velocity: 2 and 3 m/s
Microwave vacuum -Power: 200–600 W
Dragon fruit ANN
drying -Vacuum level: 3–9 kPa
-Power: 500, 1000, and 1500 W
Infrared–convective -Temperature: 40, 55, and 70 ◦ C
White Mulberry ANN and FL [163]
drying -Inlet drying air speed: 0.4, 1, and
1.6 m/s
-Temperatures: 50, 60, 70, 80, and
Mahaleb puree ANN Infrared drying [164]
90 ◦ C
-Power: 200–600 W
-Temperature: 40 to 80 ◦ C
Intermittent microwave -Air velocities: 1–2 m/s
Apple slices Real-time CVS [165]
convective drying -Pulse ratios: 2–6
-Drying rate: 0.014 to 0.000001 min−1
-Time: 27 to 244 min
-Temperature: 70 ◦ C
Hybrid hot air–infrared
Kiwifruit CVS -Thickness: 3 mm [166]
drying
-Air velocity: 0.5 or 1.5 m/s
Key: ANN, artificial neural network; CVS, computer vision system; FL, fuzzy logic.

5.8. AI in Packaging and Sorting

In the food processing industry, the organization and packaging of food products
are among the most labor-intensive and time-consuming tasks for manufacturing units.
AI-based systems can effectively manage these processes, reducing the likelihood of errors
and significantly enhancing the production efficiency of the industry. The development of
AI-based systems for sorting and packaging in the food industry is challenging due to the
variations in shape, color, and size of fruits and vegetables. To create an effective AI-driven
system, a substantial amount of data is required to properly train the model, ensuring it
performs the sorting and packaging tasks efficiently [167,168]. Numerous research teams
have developed different systems to accomplish the same task. One such system, TOMRA,
efficiently handles sorting with an impressive accuracy rate of 90%, leading to a significant
increase in production speed. Currently, most sorting and packaging tasks in the industry
are carried out by automated systems. The adoption of these technologies offers several
benefits, including faster production rates, higher-quality output, and reduced labor costs.
AI-based intelligent decision-making systems incorporate a range of tools and tech-
nologies, including high-resolution cameras, laser systems, X-ray imaging, and IR spec-
troscopy. These technologies enable a thorough analysis of various characteristics of food
products, such as fruits and vegetables, at the input stage. In contrast to conventional sys-
tems that primarily assess products based on their external appearance, advanced systems
like TOMRA have shown a 5–10% improvement in sorting and separation accuracy, specif-
ically for potatoes [169,170]. A similar issue was addressed by a Japanese company that
implemented a TensorFlow-based ML system, yielding impressive results and significant
benefits for their production line. This system also demonstrated remarkable performance
across various other food processing sectors. Additionally, each organization found that the
AI-based system operated with greater precision. The success achieved in potato processing
Foods 2025, 14, 411 29 of 38

has prompted the expansion of AI technologies to other areas within the industry, with
potential for further application across different departments of food processing.

6. Drawbacks and Limitation of AI

The development and deployment of different algorithms in AI offer unique advan-
tages and disadvantages, often tailored to specific applications and industries. Understand-
ing these benefits and limitations helps in selecting the most appropriate method for each
context. Below is an analysis of several notable algorithms (Table 3).

Table 3. Advantages and disadvantages of AI algorithms.

Algorithm Advantages Disadvantages

-High reliability and interpretability. -High initial cost due to specialized
-Quick response to input changes. requirements.
ES
-Strong in practical tasks with low error rates. -Limited vocabulary, posing communication
-Increases resource efficiency. challenges for non-experts.
-Handles vague, incomplete data well.
-Quick and simple results. -Limited to specific rules, reducing
FL -Tolerates noise, enhancing robustness. generalization.
-Flexible rule expansion. -Often depends on expert input to create rules.
-Cost-effective, improving quality and safety.
-Models complex functions accurately.
-Complex to interpret due to black-box nature.
-Resilient to noise and environmental changes.
-Longer training time, requiring specific
ANN -Learns patterns autonomously.
layer configurations.
-Generalizes well.
-Needs large, high-quality datasets.
-Cost-effective and adaptable for nonlinear problems.
-Computationally intensive, needing
-Highly effective for image-based tasks like
significant resources.
object detection.
CNN -Prone to overfitting, especially with
-Identifies complex patterns across multiple layers.
small datasets.
-Learns features automatically, reducing manual work.
-Difficult to interpret due to depth.
-Non-invasive, allowing visual assessments (size,
-High setup costs for equipment and
shape, and color) without damaging samples.
specialized software.
CVS -Useful for automated visual inspection, especially in
-Limited in low-contrast environments,
quality control.
requiring optimal lighting.
-Integrates easily with AI for accuracy.

Expert system:
ESs are AI programs designed to mimic human decision-making in specific domains.
They provide high reliability and interpretability, allowing them to perform consistently
and accurately in well-defined fields. For example, an ES can swiftly respond to input
changes, adapting to new scenarios as data evolve. This quick responsiveness makes them
suitable for time-sensitive applications such as diagnostics and resource management,
where they often outperform human operators in terms of error rates. Furthermore, ESs can
enhance the efficiency of resource use by optimizing processes based on predefined rules.
However, implementing an ES involves significant initial costs due to the specialized design
and the need for expertise in the specific domain. Additionally, ESs typically operate within
a limited vocabulary and may struggle to communicate results effectively to non-specialists.
This limited expressiveness may hinder broader application in fields that require adaptable
or user-friendly interfaces.
Fuzzy logic:
FL systems are highly effective at handling uncertain or incomplete information,
making them valuable in fields where data may lack precision, such as environmental
monitoring or control systems. One advantage of FL is its simplicity and speed, allowing
Foods 2025, 14, 411 30 of 38

it to deliver quick results without requiring extensive computations. FL systems are

also robust in noisy environments, maintaining performance despite data irregularities.
Additionally, FL is cost-effective and offers flexibility, as rule expansion allows systems
to evolve alongside changing requirements, enhancing the quality and safety of various
applications. However, FL is restricted to specific rule sets, limiting its generalizability
across different domains. These systems often rely on expert input to define comprehensive
rules, which may hinder adaptability in complex scenarios where broader contextual
understanding is necessary.
Artificial neural networks:
ANNs are well-suited for modeling complex, nonlinear relationships due to their
ability to learn autonomously from data. They excel in pattern recognition tasks, such
as image classification or NLP, as they can generalize across similar inputs. ANNs also
perform well in noisy and fluctuating environments, as they are resilient to input variability.
Furthermore, ANNs are cost-effective for nonlinear problem-solving, given their adapt-
ability across diverse applications. Despite these advantages, ANNs are often criticized for
their “black-box” nature, as their complex, multi-layered structure makes them difficult to
interpret. Understanding how an ANN reaches a decision can be challenging, especially
in critical applications that require transparency. Additionally, ANNs require significant
training time and careful selection of layer configurations, as well as large, high-quality
datasets to perform optimally, which can be resource intensive.
Convolutional neural networks:
CNNs are specialized for image-based tasks, making them highly effective for ap-
plications in object detection, image recognition, and classification. CNNs automatically
learn crucial features within images, reducing the need for manual feature engineering and
allowing the network to identify complex patterns across multiple layers of processing.
This automated feature learning enhances CNN efficiency in visual tasks. However, CNNs
demand high computational resources, making them costly and challenging to implement
without advanced hardware. Additionally, CNNs are vulnerable to overfitting, especially
when trained on small datasets, as they can become too finely tuned to specific data patterns.
Their layered structure also complicates interpretation, making it difficult to understand
how they reach decisions.
Computer vision systems:
CVSs offer a non-invasive method for assessing visual qualities like size, shape, and
color without damaging samples, making them valuable for quality control in industries
such as food production. A CVS enables automated visual inspection, which enhances
efficiency and consistency in processes that require high accuracy. Additionally, a CVS
can integrate with other AI methods to further boost accuracy, improving the precision of
assessments. The primary drawbacks of CVSs are the high initial setup costs, as they require
specialized equipment and software. Additionally, CVSs struggle in low-contrast environ-
ments, where optimal lighting is essential for reliable results. This dependency on controlled
lighting conditions limits their applicability in environments with variable lighting.
In conclusion, each algorithm has distinct advantages that make it suitable for specific
tasks and limitations that may restrict its broader application. Selecting the appropri-
ate algorithm depends on the unique demands of the application, balancing efficiency,
interpretability, and resource requirements.

7. Future Challenges
Significant advancements in food science have transformed food-based products into
nutrient-rich supplements that help protect against various diseases. AI is increasingly
being used to monitor changes in water quality and the impact of fertilizers on crop
Foods 2025, 14, 411 31 of 38

yields through the use of cameras and drones. In the production sector, AI plays a role
in reducing food waste across industries and is used in restaurants to scan food items
for their nutritional content. An important area of development involves methods for
quantifying nanomaterials in food, with AI aiding intelligent packaging by detecting
nanoscale substances in contact with food.
AI also bridges the gap between manufacturers by transferring vital information to
the cloud, creating large datasets. Automated orchard harvesting, which enables fruit
cultivation in previously unsuitable environments, helps save labor and optimize yields. AI
is poised to address fluctuating supply and demand, narrow food hygiene standards, and
improve supply chain management. Predicting the expiration dates of packaged food using
sensors presents another challenge, but it could help consumers avoid foodborne illnesses
by detecting spoilage in advance. Although the development of vendor applications is
costly and mainly targeted at larger operations, expanding these AI-driven applications
will facilitate the integration of restaurant robots in the near future.

8. Conclusions
The integration of AI across the agro-food supply chain represents a transformative
force with the potential to address critical challenges facing modern agriculture and food
systems. By leveraging state-of-the-art techniques such as ML, DL, FL, and ESs, AI en-
hances productivity, optimizes resource usage, and supports sustainable practices from
farm to fork. Applications in agriculture—spanning preproduction, production, process-
ing, and distribution—demonstrate AI’s ability to predict crop yields, improve irrigation
efficiency, detect diseases, and enhance soil management, thereby advancing precision
farming. Similarly, the food industry benefits from AI in areas such as quality control,
safety monitoring, inventory management, and waste reduction. Innovations such as CVSs,
e-noses, and advanced robotics streamline food processing, ensure safety, and mitigate
resource wastage.
The adoption of AI-driven technologies extends to emerging domains like 3D food
printing and intelligent packaging, underscoring AI’s potential to revolutionize the con-
sumer experience and reduce environmental impact. Despite its transformative capabilities,
challenges such as high implementation costs, the complexity of integrating AI with ex-
isting systems, and the need for comprehensive policies to ensure ethical and sustainable
use remain significant barriers. Future developments in AI must prioritize inclusivity,
accessibility, and cross-disciplinary collaboration to maximize their impact on global food
security and sustainability.
In conclusion, AI is no longer a supplementary tool but a critical enabler of innovation
in the agro-food system. Its strategic implementation across various stages of the food
supply chain holds promise for creating resilient, efficient, and sustainable practices capable
of feeding a growing global population while maintaining environmental balance.

Author Contributions: Conceptualization, A.A., F.A. and M.H.; methodology, A.A. and F.A.; software,
A.A. and F.A.; investigation, A.A., F.A. and M.H.; writing—original draft preparation, A.A., F.A.,
M.P. and M.H.; writing—review and editing, A.A., F.A., M.P. and M.H.; supervision, M.P. and M.H.;
funding acquisition, M.P. and M.H. All authors have read and agreed to the published version of
the manuscript.

Funding: Open Access Funding by the University of Vienna.

Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.

Data Availability Statement: The data presented in this study are available in this article.
Foods 2025, 14, 411 32 of 38

Conflicts of Interest: Author Fatemeh Aghababaei was employed by the company Aora Health.
The remaining authors declare that the research was conducted in the absence of any commercial or
financial relationships that could be construed as a potential conflict of interest.

References
1. Hrybiuk, O.; Kant, G.S. CleverCOMSRL: Implementation of an AI Computer-Aided Design System in the Context of the Cognitive Science
Paradigm for the Research Training Process In Lecture Notes in Mechanical Engineering; Springer: Cham, Switzerland, 2024; pp. 351–362.
[CrossRef]
2. Mavani, N.R.; Ali, J.M.; Othman, S.; Hussain, M.A.; Hashim, H.; Rahman, N.A. Application of Artificial Intelligence in Food
Industry—A Guideline. Food Eng. Rev. 2021, 14, 134–175. [CrossRef]
3. Ramakrishna, R.R.; Hamid, Z.A.; Zaki, W.M.D.W.; Huddin, A.B.; Mathialagan, R. Stem Cell Imaging through Convolutional
Neural Networks: Current Issues and Future Directions in Artificial Intelligence Technology. PeerJ 2020, 8, e10346. [CrossRef]
[PubMed]
4. Kawakami, E.; Tabata, J.; Yanaihara, N.; Ishikawa, T.; Koseki, K.; Iida, Y.; Saito, M.; Komazaki, H.; Shapiro, J.S.; Goto, C.; et al.
Application of Artificial Intelligence for Preoperative Diagnostic and Prognostic Prediction in Epithelial Ovarian Cancer Based on
Blood Biomarkers. Clin. Cancer Res. 2019, 25, 3006–3015. [CrossRef] [PubMed]
5. Allawi, M.F.; Jaafar, O.; Ehteram, M.; Mohamad Hamzah, F.; El-Shafie, A. Synchronizing Artificial Intelligence Models for
Operating the Dam and Reservoir System. Water Resour. Manag. 2018, 32, 3373–3389. [CrossRef]
6. Nayak, J.; Vakula, K.; Dinesh, P.; Naik, B.; Pelusi, D. Intelligent Food Processing: Journey from Artificial Neural Network to Deep
Learning. Comput. Sci. Rev. 2020, 38, 100297. [CrossRef]
7. Sharma Sagar How Artificial Intelligence Is Revolutionizing Food Processing Business? Available online: https:
//towardsdatascience.com/how-artificial-intelligence-is-revolutionizing-food-processing-business-d2a6440c0360 (accessed on 6
October 2024).
8. Funes, E.; Allouche, Y.; Beltrán, G.; Jiménez, A.; Funes, E.; Allouche, Y.; Beltrán, G.; Jiménez, A. A Review: Artificial Neural
Networks as Tool for Control Food Industry Process. J. Sens. Technol. 2015, 5, 28–43. [CrossRef]
9. Correa, D.A.; Montero Castillo, P.M.; Martelo, R.J. Neural Networks in Food Industry. Contemp. Eng. Sci. 2018, 11, 1807–1826.
[CrossRef]
10. Wang, J.; Yue, H.; Zhou, Z. An Improved Traceability System for Food Quality Assurance and Evaluation Based on Fuzzy
Classification and Neural Network. Food Control 2017, 79, 363–370. [CrossRef]
11. Sun, Q.; Zhang, M.; Mujumdar, A.S. Recent Developments of Artificial Intelligence in Drying of Fresh Food: A Review. Crit. Rev.
Food Sci. Nutr. 2019, 59, 2258–2275. [CrossRef]
12. Alizadeh-Sani, M.; Mohammadian, E.; Rhim, J.W.; Jafari, S.M. PH-Sensitive (Halochromic) Smart Packaging Films Based on
Natural Food Colorants for the Monitoring of Food Quality and Safety. Trends Food Sci. Technol. 2020, 105, 93–144. [CrossRef]
13. Halonen, N.; Pálvölgyi, P.S.; Bassani, A.; Fiorentini, C.; Nair, R.; Spigno, G.; Kordas, K. Bio-Based Smart Materials for Food
Packaging and Sensors —A Review. Front. Mater. 2020, 7, 521914. [CrossRef]
14. Bhagya Raj, G.V.S.; Dash, K.K. Comprehensive Study on Applications of Artificial Neural Network in Food Process Modeling.
Crit. Rev. Food Sci. Nutr. 2022, 62, 2756–2783. [CrossRef] [PubMed]
15. Poyatos-Racionero, E.; Ros-Lis, J.V.; Vivancos, J.L.; Martínez-Máñez, R. Recent Advances on Intelligent Packaging as Tools to
Reduce Food Waste. J. Clean. Prod. 2018, 172, 3398–3409. [CrossRef]
16. Alam, A.U.; Rathi, P.; Beshai, H.; Sarabha, G.K.; Jamal Deen, M. Fruit Quality Monitoring with Smart Packaging. Sensors 2021, 21,
1509. [CrossRef]
17. Chen, S.; Brahma, S.; Mackay, J.; Cao, C.; Aliakbarian, B. The Role of Smart Packaging System in Food Supply Chain. J. Food Sci.
2020, 85, 517–525. [CrossRef]
18. Marvin, H.; Bouzembrak, Y.; van Asselt, E.; Meijer, N.; Kleter, G.; Lorentzen, G.; Johansen, L.-H. Applicability of a Food Chain
Analysis on Aquaculture of Atlantic Salmon to Identify and Monitor Vulnerabilities and Drivers of Change for the Identification
of Emerging Risks. EFSA Support. Publ. 2019, 16, 1619E. [CrossRef]
19. Marvin, H.J.P.; Bouzembrak, Y.; van der Fels-Klerx, H.J.; Kempenaar, C.; Veerkamp, R.; Chauhan, A.; Stroosnijder, S.; Top, J.;
Simsek-Senel, G.; Vrolijk, H.; et al. Digitalisation and Artificial Intelligence for Sustainable Food Systems. Trends Food Sci. Technol.
2022, 120, 344–348. [CrossRef]
20. Van Berkum, S.; Dengerink, J.; Ruben, R. The Food Systems Approach: Sustainable Solutions for a Sufficient Supply of Healthy Food;
Wageningen Economic Research: Den Haag, The Netherlands, 2018. [CrossRef]
21. Negi, A.; Rajesh, K. A Review of AI and ML Applications for Computing Systems In Proceedings of the International Conference
on Emerging Trends in Engineering and Technology, ICETET, Nagpur, India, 1–2 November 2019. [CrossRef]
22. Nath, P.C.; Mishra, A.K.; Sharma, R.; Bhunia, B.; Mishra, B.; Tiwari, A.; Nayak, P.K.; Sharma, M.; Bhuyan, T.; Kaushal, S.; et al. Recent
Advances in Artificial Intelligence towards the Sustainable Future of Agri-Food Industry. Food Chem. 2024, 447, 138945. [CrossRef]
Foods 2025, 14, 411 33 of 38

23. Patel, G.S.; Rai, A.; Das, N.N.; Singh, R.P. Smart Agriculture: Emerging Pedagogies of Deep Learning, Machine Learning; CRC Press:
Boca Raton, FL, USA, 2021.
24. Shetty, S.H.; Shetty, S.; Singh, C.; Rao, A. Supervised Machine Learning: Algorithms and Applications. In Fundamentals and
Methods of Machine and Deep Learning: Algorithms, Tools, and Applications; Scrivener Publishing: Beverly, MA, USA, 2022; pp. 1–16.
[CrossRef]
25. Li, N.; Shepperd, M.; Guo, Y. A Systematic Review of Unsupervised Learning Techniques for Software Defect Prediction. Inf.
Softw. Technol. 2020, 122, 106287. [CrossRef]
26. Traore, B.B.; Tangara, F. Data Mining Techniques on Satellite Images for Discovery of Risk Areas. Expert. Syst. Appl. 2017, 72,
443–456. [CrossRef]
27. Mohri, M. Foundations of Machine Learning; The MIT Press: Cambridge, MA, USA, 2018.
28. Kermanidis, K.L.; Maragoudakis, M.; Krichen, M. Convolutional Neural Networks: A Survey. Computers 2023, 12, 151. [CrossRef]
29. Singh, J.; Thakur, D.; Ali, F.; Gera, T.; Kwak, K.S. Deep Feature Extraction and Classification of Android Malware Images. Sensors
2020, 20, 7013. [CrossRef] [PubMed]
30. Griffel, L.M.; Delparte, D.; Whitworth, J.; Bodily, P.; Hartley, D. Evaluation of Artificial Neural Network Performance for
Classification of Potato Plants Infected with Potato Virus Y Using Spectral Data on Multiple Varieties and Genotypes. Smart Agric.
Technol. 2023, 3, 100101. [CrossRef]
31. Sun, Q.; Zhang, M.; Yang, P. Combination of LF-NMR and BP-ANN to Monitor Water States of Typical Fruits and Vegetables
during Microwave Vacuum Drying. LWT 2019, 116, 108548. [CrossRef]
32. How Drones Can Help Manage the World’s Food Supply|World Economic Forum. Available online: https://www.weforum.org/
agenda/2018/09/how-drones-can-manage-the-food-supply-chain-and-tell-you-if-what-you-eat-is-sustainable/ (accessed on 9
October 2024).
33. Sukhadia, K.; Chaudhari, M.B. A Survey on Rice Crop Yield Prediction in India Using Improved Classification Technique. Int. J.
Sci. Res. Comput. Sci. Eng. Inf. Technol. 2019, 5, 501–507. [CrossRef]
34. Gonzalez-Fernandez, I.; Iglesias-Otero, M.A.; Esteki, M.; Moldes, O.A.; Mejuto, J.C.; Simal-Gandara, J. A Critical Review on the
Use of Artificial Neural Networks in Olive Oil Production, Characterization and Authentication. Crit. Rev. Food Sci. Nutr. 2019,
59, 1913–1926. [CrossRef]
35. Lukinac, J.; Jukić, M.; Mastanjević, K.; Lučan, M. Application of Computer Vision and Image Analysis Method in Cheese-Quality
Evaluation: A Review. Ukr. Food J. 2018, 7, 192–214. [CrossRef]
36. Hannan, M.A.; Ghani, Z.A.; Hoque, M.M.; Ker, P.J.; Hussain, A.; Mohamed, A. Fuzzy Logic Inverter Controller in Photovoltaic
Applications: Issues and Recommendations. IEEE Access 2019, 7, 24934–24955. [CrossRef]
37. Wu, X.; Du, Z.; Ma, R.; Zhang, X.; Yang, D.; Liu, H.; Zhang, Y. Qualitative and Quantitative Studies of Phthalates in Extra Virgin
Olive Oil (EVOO) by Surface-Enhanced Raman Spectroscopy (SERS) Combined with Long Short Term Memory (LSTM) Neural
Network. Food Chem. 2024, 433, 137300. [CrossRef]
38. Khairuddin, S.H.; Hasan, M.H.; Hashmani, M.A.; Azam, M.H. Generating Clustering-Based Interval Fuzzy Type-2 Triangular
and Trapezoidal Membership Functions: A Structured Literature Review. Symmetry 2021, 13, 239. [CrossRef]
39. Ali, J.A.; Hannan, M.A.; Mohamed, A.; Abdolrasol, M.G.M. Fuzzy Logic Speed Controller Optimization Approach for Induction
Motor Drive Using Backtracking Search Algorithm. Measurement 2016, 78, 49–62. [CrossRef]
40. Mazen, F.M.A.; Nashat, A.A. Ripeness Classification of Bananas Using an Artificial Neural Network. Arab. J. Sci. Eng. 2019, 44,
6901–6910. [CrossRef]
41. Pakyürek, M.; Aydin, Y.; Mikail, N. Fuzzy Logic Applications in Horticulture and a Sample Design for Juice Volume Prediction in
Pomegranate (Punica Granatum L.). Appl. Ecol. Env. Res. 2019, 17, 2449–2460. [CrossRef]
42. Duong, L.T.; Nguyen, P.T.; Di Sipio, C.; Di Ruscio, D. Automated Fruit Recognition Using EfficientNet and MixNet. Comput.
Electron. Agric. 2020, 171, 105326. [CrossRef]
43. Ahumada, O.; Villalobos, J.R. Application of Planning Models in the Agri-Food Supply Chain: A Review. Eur. J. Oper. Res. 2009,
196, 1–20. [CrossRef]
44. Ben Ayed, R.; Ennouri, K.; Ben Amar, F.; Moreau, F.; Triki, M.A.; Rebai, A. Bayesian and Phylogenic Approaches for Studying
Relationships among Table Olive Cultivars. Biochem. Genet. 2017, 55, 300–313. [CrossRef]
45. Elavarasan, D.; Vincent, D.R.; Sharma, V.; Zomaya, A.Y.; Srinivasan, K. Forecasting Yield by Integrating Agrarian Factors and
Machine Learning Models: A Survey. Comput. Electron. Agric. 2018, 155, 257–282. [CrossRef]
46. Zhang, C.; Liu, J.; Shang, J.; Cai, H. Capability of Crop Water Content for Revealing Variability of Winter Wheat Grain Yield and
Soil Moisture under Limited Irrigation. Sci. Total Environ. 2018, 631–632, 677–687. [CrossRef]
47. Morellos, A.; Pantazi, X.E.; Moshou, D.; Alexandridis, T.; Whetton, R.; Tziotzios, G.; Wiebensohn, J.; Bill, R.; Mouazen, A.M.
Machine Learning Based Prediction of Soil Total Nitrogen, Organic Carbon and Moisture Content by Using VIS-NIR Spectroscopy.
Biosyst. Eng. 2016, 152, 104–116. [CrossRef]
Foods 2025, 14, 411 34 of 38

48. Kumar, R.; Singh, M.P.; Kumar, P.; Singh, J.P. Crop Selection Method to Maximize Crop Yield Rate Using Machine Learning
Technique. In Proceedings of the 2015 International Conference on Smart Technologies and Management for Computing,
Communication, Controls, Energy and Materials, ICSTM, Avadi, India, 6–8 May 2015; pp. 138–145. [CrossRef]
49. Goap, A.; Sharma, D.; Shukla, A.K.; Rama Krishna, C. An IoT Based Smart Irrigation Management System Using Machine
Learning and Open Source Technologies. Comput. Electron. Agric. 2018, 155, 41–49. [CrossRef]
50. Arvind, G.; Athira, V.G.; Haripriya, H.; Rani, R.A.; Aravind, S. Automated Irrigation with Advanced Seed Germination and
Pest Control. In Proceedings of the 2017 IEEE Technological Innovations in ICT for Agriculture and Rural Development, TIAR,
Chennai, India, 7–8 April 2017; pp. 64–67. [CrossRef]
51. Cruz, J.R.D.; Baldovino, R.G.; Bandala, A.A.; Dadios, E.P. Water Usage Optimization of Smart Farm Automated Irrigation System
Using Artificial Neural Network. In Proceedings of the 2017 5th International Conference on Information and Communication
Technology, ICoIC7, Melaka, Malaysia, 17–19 May 2017. [CrossRef]
52. Choudhary, S.; Gaurav, V.; Singh, A.; Agarwal, S. Autonomous Crop Irrigation System Using Artificial Intelligence. Int. J. Eng.
Adv. Technol. 2019, 8, 46–51. [CrossRef]
53. Saggi, M.K.; Jain, S. Reference Evapotranspiration Estimation and Modeling of the Punjab Northern India Using Deep Learning.
Comput. Electron. Agric. 2019, 156, 387–398. [CrossRef]
54. Singh, A.; Shukla, N.; Mishra, N. Social Media Data Analytics to Improve Supply Chain Management in Food Industries. Transp.
Res. E Logist. Transp. Rev. 2018, 114, 398–415. [CrossRef]
55. Kamilaris, A.; Prenafeta-Boldú, F.X. Deep Learning in Agriculture: A Survey. Comput. Electron. Agric. 2018, 147, 70–90. [CrossRef]
56. Liakos, K.G.; Busato, P.; Moshou, D.; Pearson, S.; Bochtis, D. Machine Learning in Agriculture: A Review. Sensors 2018, 18, 2674.
[CrossRef]
57. Chlingaryan, A.; Sukkarieh, S.; Whelan, B. Machine Learning Approaches for Crop Yield Prediction and Nitrogen Status
Estimation in Precision Agriculture: A Review. Comput. Electron. Agric. 2018, 151, 61–69. [CrossRef]
58. Sadgrove, E.J.; Falzon, G.; Miron, D.; Lamb, D.W. Real-Time Object Detection in Agricultural/Remote Environments Using the
Multiple-Expert Colour Feature Extreme Learning Machine (MEC-ELM). Comput. Ind. 2018, 98, 183–191. [CrossRef]
59. Gao, Z.; Shao, Y.; Xuan, G.; Wang, Y.; Liu, Y.; Han, X. Real-Time Hyperspectral Imaging for the in-Field Estimation of Strawberry
Ripeness with Deep Learning. Artif. Intell. Agric. 2020, 4, 31–38. [CrossRef]
60. Ma, X.; Wang, S.; Bai, Q. Coordination of Production Scheduling and Vehicle Routing Problems for Perishable Food Products. Int.
J. Internet Manuf. Serv. 2019, 6, 79–96. [CrossRef]
61. Arora, M.; Mangipudi, P. A Computer Vision-Based Method for Classification of Red Meat Quality After Nitrosamine Appendage.
Int. J. Comput. Intell. Appl. 2021, 20, 2150005. [CrossRef]
62. Farah, J.S.; Cavalcanti, R.N.; Guimarães, J.T.; Balthazar, C.F.; Coimbra, P.T.; Pimentel, T.C.; Esmerino, E.A.; Duarte, M.C.K.H.;
Freitas, M.Q.; Granato, D.; et al. Differential Scanning Calorimetry Coupled with Machine Learning Technique: An Effective
Approach to Determine the Milk Authenticity. Food Control 2021, 121, 107585. [CrossRef]
63. Dolgui, A.; Tiwari, M.K.; Sinjana, Y.; Kumar, S.K.; Son, Y.J. Optimising Integrated Inventory Policy for Perishable Items in a
Multi-Stage Supply Chain. Int. J. Prod. Res. 2018, 56, 902–925. [CrossRef]
64. Krishna, K.R. Push Button Agriculture: Robotics, Drones, Satellite-Guided Soil and Crop Management; Apple Acedemic Press: Oakvile,
ON, Canada, 2016. Available online: https://books.google.es/books?hl=en&lr=&id=NlmzDAAAQBAJ&oi=fnd&pg=PP1
&dq=Push+Button+Agriculture:+Robotics,+Drones,++Satellite-Guided+Soil+and+Crop+Management&ots=TMkkjNnZ8O&sig=
Nuhzq_elyeki3EMkr-HK26PmikY&redir_esc=y#v=onepage&q=Push%20Button%20Agriculture:%20Robotics,%20Drones,%20%
20Satellite-Guided%20Soil%20and%20Crop%20Management&f=false (accessed on 9 October 2024).
65. Ravensberg, S. GIS in Agriculture—Integrate Sustainability. Available online: https://www.integratesustainability.com.au/2018
/11/23/gis-in-agriculture/ (accessed on 9 October 2024).
66. Zhao, G.; Liu, S.; Lopez, C.; Lu, H.; Elgueta, S.; Chen, H.; Boshkoska, B.M. Blockchain Technology in Agri-Food Value Chain
Management: A Synthesis of Applications, Challenges and Future Research Directions. Comput. Ind. 2019, 109, 83–99. [CrossRef]
67. Miranda, J.; Ponce, P.; Molina, A.; Wright, P. Sensing, Smart and Sustainable Technologies for Agri-Food 4.0. Comput. Ind. 2019,
108, 21–36. [CrossRef]
68. Bacco, M.; Barsocchi, P.; Ferro, E.; Gotta, A.; Ruggeri, M. The Digitisation of Agriculture: A Survey of Research Activities on
Smart Farming. Array 2019, 3–4, 100009. [CrossRef]
69. O’Grady, M.J.; Langton, D.; O’Hare, G.M.P. Edge Computing: A Tractable Model for Smart Agriculture? Artif. Intell. Agric. 2019,
3, 42–51. [CrossRef]
70. Jha, K.; Doshi, A.; Patel, P.; Shah, M. A Comprehensive Review on Automation in Agriculture Using Artificial Intelligence. Artif.
Intell. Agric. 2019, 2, 1–12. [CrossRef]
71. Kootstra, G.; Wang, X.; Blok, P.M.; Hemming, J.; van Henten, E. Selective Harvesting Robotics: Current Research, Trends, and
Future Directions. Curr. Robot. Rep. 2021, 2, 95–104. [CrossRef]
Foods 2025, 14, 411 35 of 38

72. Drones and Precision Agriculture: The Future of Farming. Available online: https://www.microdrones.com/en/content/drones-
and-precision-agriculture-the-future-of-farming/ (accessed on 9 October 2024).
73. Indian Farming with Drones|Equinox’s Drones. Available online: https://equinoxsdrones.com/importance-of-drone-
technology-in-indian-agriculture-farming/ (accessed on 9 October 2024).
74. Huuskonen, J.; Oksanen, T. Soil Sampling with Drones and Augmented Reality in Precision Agriculture. Comput. Electron. Agric.
2018, 154, 25–35. [CrossRef]
75. Robots Agricoles, Drones et IA: 2020–2040: Technologies, Marchés et Joueurs: IDTechEx. Available online: https://www.idtechex.
com/fr/research-report/agricultural-robots-drones-and-ai-2020-2040-technologies-markets-and-players/749 (accessed on 9
October 2024).
76. Krishna, K.R. Agricultural Drones: A Peaceful Pursuit; Apple Academic Press: Oakvile, ON, Canada, 2021; ISBN 9781774636428.
77. Agriculture Drone Market Size, Share|Global Report [2032]. Available online: https://www.fortunebusinessinsights.com/
agriculture-drones-market-102589 (accessed on 9 October 2024).
78. Ben Ayed, R.; Hanana, M. Artificial Intelligence to Improve the Food and Agriculture Sector. J. Food Qual. 2021, 2021, 5584754.
[CrossRef]
79. Addanki, M.; Patra, P.; Kandra, P. Recent Advances and Applications of Artificial Intelligence and Related Technologies in the
Food Industry. Appl. Food Res. 2022, 2, 100126. [CrossRef]
80. How, M.L.; Chan, Y.J.; Cheah, S.M. Predictive Insights for Improving the Resilience of Global Food Security Using Artificial
Intelligence. Sustainability 2020, 12, 6272. [CrossRef]
81. Kleineidam, J. Fields of Action for Designing Measures to Avoid Food Losses in Logistics Networks. Sustainability 2020, 12, 6093.
[CrossRef]
82. Cárdenas-Pérez, S.; Chanona-Pérez, J.; Méndez-Méndez, J.V.; Calderón-Domínguez, G.; López-Santiago, R.; Perea-Flores, M.J.;
Arzate-Vázquez, I. Evaluation of the Ripening Stages of Apple (Golden Delicious) by Means of Computer Vision System. Biosyst.
Eng. 2017, 159, 46–58. [CrossRef]
83. De Oliveira, E.M.; Leme, D.S.; Barbosa, B.H.G.; Rodarte, M.P.; Alvarenga Pereira, R.G.F. A Computer Vision System for Coffee
Beans Classification Based on Computational Intelligence Techniques. J. Food Eng. 2016, 171, 22–27. [CrossRef]
84. Sun, X.; Young, J.; Liu, J.H.; Bachmeier, L.; Somers, R.M.; Chen, K.J.; Newman, D. Prediction of Pork Color Attributes Using
Computer Vision System. Meat Sci. 2016, 113, 62–64. [CrossRef]
85. Leme, D.S.; da Silva, S.A.; Barbosa, B.H.G.; Borém, F.M.; Pereira, R.G.F.A. Recognition of Coffee Roasting Degree Using a
Computer Vision System. Comput. Electron. Agric. 2019, 156, 312–317. [CrossRef]
86. Chmiel, M.; Słowiński, M. The Use of Computer Vision System to Detect Pork Defects. LWT 2016, 73, 473–480. [CrossRef]
87. Cavallo, D.P.; Cefola, M.; Pace, B.; Logrieco, A.F.; Attolico, G. Non-Destructive and Contactless Quality Evaluation of Table
Grapes by a Computer Vision System. Comput. Electron. Agric. 2019, 156, 558–564. [CrossRef]
88. Bakhshipour, A.; Zareiforoush, H.; Bagheri, I. Application of Decision Trees and Fuzzy Inference System for Quality Classification
and Modeling of Black and Green Tea Based on Visual Features. J. Food Meas. Charact. 2020, 14, 1402–1416. [CrossRef]
89. Koklu, M.; Ozkan, I.A. Multiclass Classification of Dry Beans Using Computer Vision and Machine Learning Techniques. Comput.
Electron. Agric. 2020, 174, 105507. [CrossRef]
90. Lopes, J.F.; Ludwig, L.; Barbin, D.F.; Grossmann, M.V.E.; Barbon, S. Computer Vision Classification of Barley Flour Based on
Spatial Pyramid Partition Ensemble. Sensors 2019, 19, 2953. [CrossRef] [PubMed]
91. Villaseñor-Aguilar, M.J.; Bravo-Sánchez, M.G.; Padilla-Medina, J.A.; Vázquez-Vera, J.L.; Guevara-González, R.G.; García-
Rodríguez, F.J.; Barranco-Gutiérrez, A.I. A Maturity Estimation of Bell Pepper (Capsicum Annuum L.) by Artificial Vision System
for Quality Control. Appl. Sci. 2020, 10, 5097. [CrossRef]
92. Fan, S.; Li, J.; Zhang, Y.; Tian, X.; Wang, Q.; He, X.; Zhang, C.; Huang, W. On Line Detection of Defective Apples Using Computer
Vision System Combined with Deep Learning Methods. J. Food Eng. 2020, 286, 110102. [CrossRef]
93. Gonzalez Viejo, C.; Torrico, D.D.; Dunshea, F.R.; Fuentes, S. Development of Artificial Neural Network Models to Assess Beer
Acceptability Based on Sensory Properties Using a Robotic Pourer: A Comparative Model Approach to Achieve an Artificial
Intelligence System. Beverages 2019, 5, 33. [CrossRef]
94. Phate, V.R.; Malmathanraj, R.; Palanisamy, P. Clustered ANFIS Weighing Models for Sweet Lime (Citrus Limetta) Using Computer
Vision System. J. Food Process Eng. 2019, 42, e13160. [CrossRef]
95. Nadim, M.; Ahmadifar, H.; Mashkinmojeh, M.; Yamaghani, M.R. Application of Image Processing Techniques for Quality Control
of Mushroom. Casp. J. Health Res. 2019, 4, 72–75. [CrossRef]
96. Castro, W.; Oblitas, J.; De-La-Torre, M.; Cotrina, C.; Bazan, K.; Avila-George, H. Classification of Cape Gooseberry Fruit According
to Its Level of Ripeness Using Machine Learning Techniques and Different Color Spaces. IEEE Access 2019, 7, 27389–27400.
[CrossRef]
97. Andrew, J.; Garcia, A.; Arboleda, E.R.; Galas, E.M. Identification of Visually Similar Vegetable Seeds Using Image Processing And
Fuzzy Logic. Int. J. Sci. Technol. Res. 2020, 9, 2.
Foods 2025, 14, 411 36 of 38

98. Sun, X.; Young, J.; Liu, J.H.; Newman, D. Prediction of Pork Loin Quality Using Online Computer Vision System and Artificial
Intelligence Model. Meat Sci. 2018, 140, 72–77. [CrossRef]
99. Wan, P.; Toudeshki, A.; Tan, H.; Ehsani, R. A Methodology for Fresh Tomato Maturity Detection Using Computer Vision. Comput.
Electron. Agric. 2018, 146, 43–50. [CrossRef]
100. Sidehabi, S.W.; Suyuti, A.; Areni, I.S.; Nurtanio, I. The Development of Machine Vision System for Sorting Passion Fruit Using
Multi-Class Support Vector Machine. J. Eng. Sci. Technol. Rev. 2018, 11, 178–184. [CrossRef]
101. Rodríguez-Espíndola, O.; Chowdhury, S.; Beltagui, A.; Albores, P. The Potential of Emergent Disruptive Technologies for
Humanitarian Supply Chains: The Integration of Blockchain, Artificial Intelligence and 3D Printing. Int. J. Prod. Res. 2020, 58,
4610–4630. [CrossRef]
102. Delcour, N.; Duncan, L.; Frahm, S.; Lancaster, P.; Vann, L. Estimation of Technology Convergence by 2035. Available online:
https://apps.dtic.mil/sti/citations/AD1109853 (accessed on 6 October 2024).
103. Ronsivalle, V.; Ruiz, F.; Lo Giudice, A.; Carli, E.; Venezia, P.; Isola, G.; Leonardi, R.; Mummolo, S. From Reverse Engineering
Software to CAD-CAM Systems: How Digital Environment Has Influenced the Clinical Applications in Modern Dentistry and
Orthodontics. Appl. Sci. 2023, 13, 4986. [CrossRef]
104. Lanaro, M.; Desselle, M.R.; Woodruff, M.A. 3D Printing Chocolate: Properties of Formulations for Extrusion, Sintering, Binding
and Ink Jetting. In Fundamentals of 3D Food Printing and Applications; Academic Press: New York, NY, USA, 2019; pp. 151–173.
[CrossRef]
105. Beckett, S.T. Industrial Chocolate Manufacture and Use, 4th ed.; Wiley-Blackwell: Hoboken, NJ, USA, 2009; p. 688. [CrossRef]
106. Dankar, I.; Haddarah, A.; Omar, F.E.L.; Sepulcre, F.; Pujolà, M. 3D Printing Technology: The New Era for Food Customization and
Elaboration. Trends Food Sci. Technol. 2018, 75, 231–242. [CrossRef]
107. Guo, C.; Zhang, M.; Bhandari, B. Model Building and Slicing in Food 3D Printing Processes: A Review. Compr. Rev. Food Sci. Food
Saf. 2019, 18, 1052–1069. [CrossRef]
108. Sakhti, M.V.; Priya, R.M.; Krisnaraj, N.; Prabhakar, E. A Novel Artificial Intelligent System for Milk Conservation Using Wireless
Sensor Networks. Bonfring Int. J. Netw. Technol. Appl. 2012, 1, 7–13. [CrossRef]
109. Meshram, B.; Suvartan, R.; Shaikh, A. Robotics: An Emerging Technology in Dairy. and Food Industry: Review. Int. J. Chem. Stud.
2018, 6, 440–449.
110. Khoroshailo, T.A.; Kozub, Y.A. Robotization in the Production of Dairy, Meat and Fish Products. J. Phys. Conf. Ser. 2020, 1515,
022007. [CrossRef]
111. Yanai, K.; Kawano, Y. Food Image Recognition Using Deep Convolutional Network with Pre-Training and Fine-Tuning. In
Proceedings of the 2015 IEEE International Conference on Multimedia and Expo. Workshops, ICMEW, Turin, Italy, 29 June–3 July
2015. [CrossRef]
112. Kakani, V.; Nguyen, V.H.; Kumar, B.P.; Kim, H.; Pasupuleti, V.R. A Critical Review on Computer Vision and Artificial Intelligence
in Food Industry. J. Agric. Food Res. 2020, 2, 100033. [CrossRef]
113. Fedorova, E.; Darbasov, V.; Okhlopkov, M. The Role of Agricultural Economists in Study on Problems Related to Regional Food
Safety. E3S Web Conf. 2020, 176, 05011. [CrossRef]
114. Kumar, I.; Rawat, J.; Mohd, N.; Husain, S. Opportunities of Artificial Intelligence and Machine Learning in the Food
Industry. J. Food Qual. 2021, 2021, 4535567. [CrossRef]
115. Yu, X.; Lin, Y.; Wu, H. Targeted Next-Generation Sequencing Identifies Separate Causes of Hearing Loss in One Deaf Family and
Variable Clinical Manifestations for the p.R161C Mutation in SOX10. Neural Plast. 2020, 2020, 8860837. [CrossRef] [PubMed]
116. Filimonau, V.; Todorova, E.; Mzembe, A.; Sauer, L.; Yankholmes, A. A Comparative Study of Food Waste Management in Full
Service Restaurants of the United Kingdom and the Netherlands. J. Clean. Prod. 2020, 258, 120775. [CrossRef]
117. Viejo, C.G.; Torrico, D.D.; Dunshea, F.R.; Fuentes, S. Emerging Technologies Based on Artificial Intelligence to Assess the Quality
and Consumer Preference of Beverages. Beverages 2019, 5, 62. [CrossRef]
118. Vassileva, S.; Mileva, S. Ai-Based Software Tools for Beer Brewing Monitoring and Control. Biotechnol. Biotechnol. Equip. 2014, 24,
1936–1939. [CrossRef]
119. OIV Descriptor List for Grape Varieties and Vitis Species. Available online: http://www.bioversityinternational.org/publications/
pubfile.asp?id_pub=124 (accessed on 1 May 2023).
120. Palacios, F.; Diago, M.P.; Tardaguila, J. A Non-Invasive Method Based on Computer Vision for Grapevine Cluster Compactness
Assessment Using a Mobile Sensing Platform under Field Conditions. Sensors 2019, 19, 3799. [CrossRef]
121. Hafiz, R.; Haque, M.R.; Rakshit, A.; Uddin, M.S. Image-Based Soft Drink Type Classification and Dietary Assessment System
Using Deep Convolutional Neural Network with Transfer Learning. J. King Saud. Univ. Comput. Inf. Sci. 2022, 34, 1775–1784.
[CrossRef]
122. Thazin, Y.; Pobkrut, T.; Kerdcharoen, T. Prediction of Acidity Levels of Fresh Roasted Coffees Using E-Nose and Artificial Neural
Network. In Proceedings of the 2018 10th International Conference on Knowledge and Smart Technology: Cybernetics in the
Next Decades, KST, Chiang Mai, Thailand, 31 January–3 February 2018; pp. 210–215. [CrossRef]
Foods 2025, 14, 411 37 of 38

123. Nashat, S.; Abdullah, M.Z. Quality Evaluation of Bakery Products. Computer Vision Technology for Food Quality Evaluation; 2nd ed.;
Academic Press: New York, NY, USA, 2016; pp. 525–589. [CrossRef]
124. Lingle, T.R.; Menon, S.N. Cupping and Grading—Discovering Character and Quality. In The Craft and Science of Coffee; Academic
Press: New York, NY, USA, 2017; pp. 181–203. [CrossRef]
125. Tan, J.; Xu, J. Applications of Electronic Nose (e-Nose) and Electronic Tongue (e-Tongue) in Food Quality-Related Properties
Determination: A Review. Artif. Intell. Agric. 2020, 4, 104–115. [CrossRef]
126. Gao, L.; Bian, M.; Mi, R.; Hu, X.; Wu, J. Quality Identification and Evaluation of Pu-Erh Teas of Different Grade Levels and Various
Ages through Sensory Evaluation and Instrumental Analysis. Int. J. Food Sci. Technol. 2016, 51, 1338–1348. [CrossRef]
127. Banerjee (Roy), R.; Tudu, B.; Bandyopadhyay, R.; Bhattacharyya, N. Application of Electronic Nose and Tongue for Beverage
Quality Evaluation. Eng. Tools Beverage Ind. 2019, 3, 229–254. [CrossRef]
128. Pizzaia, J.P.L.; Salcides, I.R.; de Almeida, G.M.; Contarato, R.; de Almeida, R. Arabica Coffee Samples Classification Using a
Multilayer Perceptron Neural Network. In Proceedings of the 2018 13th IEEE International Conference on Industry Applications,
INDUSCON, Sao Paulo, Brazil, 2–14 November 2018; pp. 80–84. [CrossRef]
129. Valley, P. User Acceptance of Baking Automation: An Extension of the Unified Theory of Acceptance and Use of Technology (UTAUT) to
Manufacturing; ProQuest: Ann Arbor, MI, USA, 2017.
130. Khorolskyi, V.; Yermak, S.; Bavyko, O.; Korenets, Y.; Riabykina, N. Technological Complex of Automated Control and Management
of Water Purification and Bread Production with Robotic Technologic Intensifiers. J. Hyg. Eng. Des. 2018, 25, 112–120.
131. Karyn, L.N. Sourdough by Science: Understanding Bread Making for Successful Baking. Available online: https:
//books.google.es/books?hl=en&lr=&id=3EM3EAAAQBAJ&oi=fnd&pg=PT21&dq=Overall,+for+the+complete+baking+
process+starting+from+mixing+and+till+packaging+requires+about+3+hours.+&ots=JI4I3OP18C&sig=ozlmPUMn46gr9
0ulZ2qcYFwg16I&redir_esc=y#v=onepage&q&f=false (accessed on 6 October 2024).
132. Chen, J.; Zhang, M.; Xu, B.; Sun, J.; Mujumdar, A.S. Artificial Intelligence Assisted Technologies for Controlling the Drying of
Fruits and Vegetables Using Physical Fields: A Review. Trends Food Sci. Technol. 2020, 105, 251–260. [CrossRef]
133. Chobot, M.; Kozłowska, M.; Ignaczak, A.; Kowalska, H. Development of Drying and Roasting Processes for the Production of
Plant-Based pro-Healthy Snacks in the Light of Nutritional Trends and Sustainable Techniques. Trends Food Sci. Technol. 2024, 149,
104553. [CrossRef]
134. Monton, C.; Luprasong, C.; Charoenchai, L. Acceleration of Turmeric Drying Using Convection and Microwave-Assisted Drying
Technique: An Optimization Approach. J. Food Process Preserv. 2019, 43, e14096. [CrossRef]
135. Zhang, M.; Tang, J.; Mujumdar, A.S.; Wang, S. Trends in Microwave-Related Drying of Fruits and Vegetables. Trends Food Sci.
Technol. 2006, 17, 524–534. [CrossRef]
136. Lv, W.; Zhang, M.; Bhandari, B.; Li, L.; Wang, Y. Smart NMR Method of Measurement of Moisture Content of Vegetables During
Microwave Vacuum Drying. Food Bioprocess Technol. 2017, 10, 2251–2260. [CrossRef]
137. Chen, Y.; Martynenko, A. Computer Vision for Real-Time Measurements of Shrinkage and Color Changes in Blueberry Convective
Drying. Dry. Technol. 2013, 31, 1114–1123. [CrossRef]
138. Jiao, Y.; Tang, J.; Wang, Y.; Koral, T.L. Radio-Frequency Applications for Food Processing and Safety. Annu. Rev. Food Sci. Technol.
2018, 9, 105–127. [CrossRef]
139. Zhou, X.; Xu, R.; Zhang, B.; Pei, S.; Liu, Q.; Ramaswamy, H.S.; Wang, S. Radio Frequency-Vacuum Drying of Kiwifruits: Kinetics,
Uniformity, and Product Quality. Food Bioprocess Technol. 2018, 11, 2094–2109. [CrossRef]
140. Pawar, S.B.; Pratape, V.M. Fundamentals of Infrared Heating and Its Application in Drying of Food Materials: A Review. J. Food
Process Eng. 2017, 40, e12308. [CrossRef]
141. Adak, N.; Heybeli, N.; Ertekin, C. Infrared Drying of Strawberry. Food Chem. 2017, 219, 109–116. [CrossRef] [PubMed]
142. Mihindukulasuriya, S.D.F.; Jayasuriya, H.P.W. Drying of Chilli in a Combined Infrared and Hot Air Rotary Dryer. J. Food Sci.
Technol. 2015, 52, 4895–4904. [CrossRef] [PubMed]
143. Nadian, M.H.; Abbaspour-Fard, M.H.; Martynenko, A.; Golzarian, M.R. An Intelligent Integrated Control of Hybrid Hot
Air-Infrared Dryer Based on Fuzzy Logic and Computer Vision System. Comput. Electron. Agric. 2017, 137, 138–149. [CrossRef]
144. Huang, L.; Zhang, M. Trends in Development of Dried Vegetable Products as Snacks. Dry. Technol. 2012, 30, 448–461. [CrossRef]
145. Xu, B.; Yuan, J.; Wang, L.; Lu, F.; Wei, B.; Azam, R.S.M.; Ren, X.; Zhou, C.; Ma, H.; Bhandari, B. Effect of Multi-Frequency Power
Ultrasound (MFPU) Treatment on Enzyme Hydrolysis of Casein. Ultrason. Sonochem 2020, 63, 104930. [CrossRef]
146. Bromberger Soquetta, M.; Schmaltz, S.; Wesz Righes, F.; Salvalaggio, R.; de Marsillac Terra, L. Effects of Pretreatment Ultrasound
Bath and Ultrasonic Probe, in Osmotic Dehydration, in the Kinetics of Oven Drying and the Physicochemical Properties of Beet
Snacks. J. Food Process Preserv. 2018, 42, e13393. [CrossRef]
147. Wang, L.; Xu, B.; Wei, B.; Zeng, R. Low Frequency Ultrasound Pretreatment of Carrot Slices: Effect on the Moisture Migration and
Quality Attributes by Intermediate-Wave Infrared Radiation Drying. Ultrason. Sonochem 2018, 40, 619–628. [CrossRef]
148. Rodríguez, Ó.; Santacatalina, J.V.; Simal, S.; Garcia-Perez, J.V.; Femenia, A.; Rosselló, C. Influence of Power Ultrasound Application
on Drying Kinetics of Apple and Its Antioxidant and Microstructural Properties. J. Food Eng. 2014, 129, 21–29. [CrossRef]
Foods 2025, 14, 411 38 of 38

149. Pu, Y.Y.; Sun, D.W. Prediction of Moisture Content Uniformity of Microwave-Vacuum Dried Mangoes as Affected by Different
Shapes Using NIR Hyperspectral Imaging. Innov. Food Sci. Emerg. Technol. 2016, 33, 348–356. [CrossRef]
150. Taghinezhad, E.; Kaveh, M.; Jahanbakhshi, A.; Golpour, I. Use of Artificial Intelligence for the Estimation of Effective Moisture
Diffusivity, Specific Energy Consumption, Color and Shrinkage in Quince Drying. J. Food Process Eng. 2020, 43, e13358. [CrossRef]
151. Dai, A.; Zhou, X.; Dang, H.; Sun, M.; Wu, Z. Intelligent Modeling Method for a Combined Radiation-Convection Grain Dryer:
A Support Vector Regression Algorithm Based on an Improved Particle Swarm Optimization Algorithm. IEEE Access 2018, 6,
14285–14297. [CrossRef]
152. Tarafdar, A.; Jothi, N.; Kaur, B.P. Mathematical and Artificial Neural Network Modeling for Vacuum Drying Kinetics of Moringa
Olifera Leaves Followed by Determination of Energy Consumption and Mass Transfer Parameters. J. Appl. Res. Med. Aromat.
Plants 2021, 24, 100306. [CrossRef]
153. Fabani, M.P.; Capossio, J.P.; Román, M.C.; Zhu, W.; Rodriguez, R.; Mazza, G. Producing Non-Traditional Flour from Watermelon
Rind Pomace: Artificial Neural Network (ANN) Modeling of the Drying Process. J. Env. Manag. 2021, 281, 111915. [CrossRef]
[PubMed]
154. Kalsi, B.S.; Singh, S.; Alam, M.S.; Bhatia, S. Microwave Drying Modelling of Stevia Rebaudiana Leaves Using Artificial Neural
Network and Its Effect on Color and Biochemical Attributes. J. Food Qual. 2023, 2023, 2811491. [CrossRef]
155. Kalathingal, M.S.H.; Basak, S.; Mitra, J. Artificial Neural Network Modeling and Genetic Algorithm Optimization of Process
Parameters in Fluidized Bed Drying of Green Tea Leaves. J. Food Process Eng. 2020, 43, e13128. [CrossRef]
156. Liu, Z.L.; Bai, J.W.; Wang, S.X.; Meng, J.S.; Wang, H.; Yu, X.L.; Gao, Z.J.; Xiao, H.W. Prediction of Energy and Exergy of Mushroom
Slices Drying in Hot Air Impingement Dryer by Artificial Neural Network. Dry. Technol. 2020, 38, 1959–1970. [CrossRef]
157. Yang, T.; Zheng, X.; Vidyarthi, S.K.; Xiao, H.; Yao, X.; Li, Y.; Zang, Y.; Zhang, J. Artificial Neural Network Modeling and Genetic
Algorithm Multiobjective Optimization of Process of Drying-Assisted Walnut Breaking. Foods 2023, 12, 1897. [CrossRef]
158. Chokphoemphun, S.; Hongkong, S.; Chokphoemphun, S. Evaluation of Drying Behavior and Characteristics of Potato Slices in
Multi–Stage Convective Cabinet Dryer: Application of Artificial Neural Network. Inf. Process. Agric. 2024, 11, 457–475. [CrossRef]
159. Rasooli Sharabiani, V.; Kaveh, M.; Abdi, R.; Szymanek, M.; Tanaś, W. Estimation of Moisture Ratio for Apple Drying by Convective
and Microwave Methods Using Artificial Neural Network Modeling. Sci. Rep. 2021, 11, 9155. [CrossRef]
160. Omari, A.; Behroozi-Khazaei, N.; Sharifian, F. Drying Kinetic and Artificial Neural Network Modeling of Mushroom Drying
Process in Microwave-Hot Air Dryer. J. Food Process Eng. 2018, 41, e12849. [CrossRef]
161. Bai, J.W.; Xiao, H.W.; Ma, H.L.; Zhou, C.S. Artificial Neural Network Modeling of Drying Kinetics and Color Changes of Ginkgo
Biloba Seeds during Microwave Drying Process. J. Food Qual. 2018, 2018, 3278595. [CrossRef]
162. Jafari, S.M.; Ganje, M.; Dehnad, D.; Ghanbari, V. Mathematical, Fuzzy Logic and Artificial Neural Network Modeling Techniques
to Predict Drying Kinetics of Onion. J. Food Process Preserv. 2016, 40, 329–339. [CrossRef]
163. Golpour, I.; Kaveh, M.; Amiri Chayjan, R.; Guiné, R.P.F. Optimization of Infrared-Convective Drying of White Mulberry Fruit
Using Response Surface Methodology and Development of a Predictive Model through Artificial Neural Network. Int. J. Fruit.
Sci. 2020, 20 (Suppl. S2), S1015–S1035. [CrossRef]
164. Isleroglu, H.; Beyhan, S. Intelligent Models Based Nonlinear Modeling for Infrared Drying of Mahaleb Puree. J. Food Process Eng.
2018, 41, e12912. [CrossRef]
165. Aghilinategh, N.; Rafiee, S.; Hosseinpour, S.; Omid, M.; Mohtasebi, S.S. Real-Time Color Change Monitoring of Apple Slices
Using Image Processing during Intermittent Microwave Convective Drying. Food Sci. Technol. Int. 2016, 22, 634–646. [CrossRef]
166. Nadian, M.H.; Abbaspour-Fard, M.H.; Sadrnia, H.; Golzarian, M.R.; Tabasizadeh, M.; Martynenko, A. Improvement of Kiwifruit
Drying Using Computer Vision System (CVS) and ALM Clustering Method. Dry. Technol. 2017, 35, 709–723. [CrossRef]
167. Kapur, P.K.; Singh, O.; Khatri, S.K.; Verma, A.K. (Eds.) Strategic System Assurance and Business Analytics; Asset Analytics Springer
Singapore: Singapore, 2020; ISBN 978-981-15-3646-5.
168. Dewi, T.; Risma, P.; Oktarina, Y. Fruit Sorting Robot Based on Color and Size for an Agricultural Product Packaging System. Bull.
Electr. Eng. Inform. 2020, 9, 1438–1445. [CrossRef]
169. Project-Based Learning at WPI | PBL in Higher Education. Available online: https://www.wpi.edu/project-based-learning
(accessed on 12 October 2024).
170. Оnishchuk, M.O. Opto-Mechanical Sorting of Municipal Solid Waste; Vinnytsia National Technical University: Vinnytsia Oblast,
Ukraine, 2018.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.