Volume 7, Issue 6, June – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Soil Classification and Crop Recommendation System

Ashwin John, Denzil Davis, Athulya Merin C Tom, Dibin Davis, Jasmi Davis*
Asst. Professor* Computer Science and Engineering Department
Sahrdaya College of Engineering and Technology Thrissur, India

Abstract:- In this project we will be making a Soil the choice of key options for a selected region and keeps
Classification and Crop Recommendation System. This the crop prediction method perpetually upgraded. This
appplication will help the farmers to test the quality of work proposes a unique feature choice approach known as
the soil for the cultivation, so the farmers no need of changed algorithmic feature elimination to pick out
going to the laboratories for testing the soil. By checking acceptable options from an information set for crop
this we can find which crop can give more yield. With prediction.The planned changed algorithmic feature
the help of a smartphone the farmers can test the soil by elimination technique selects and ranks salient options
themselves. We are implementing this system by employing a ranking method. The experimental results
applying machine learning algorithm. The models are show that the changed algorithmic feature elimination
trained on the basis of a large dataset, so it will increase methodology selects the most correct options, whereas the
the accuracy of the model. material technique helps accurately predict a suitable crop.

Keywords:- Quality, Testing, Machine Leraning Algorithm. In the Paper “ Smartphone-Based Spectrometric
Analyzer for Accurate Estimation of pH Value in Soil “ [3]
I. INTRODUCTION - It demonstrates a cheap, compact, and hand-held
smartphone-based sensing tool for correct estimation of pH
Agriculture is that the major supply for living for the scale values of agricultural farmlands. we tend to develop a
folks of Asian nation. Agriculture analysis is that the major qualitative analysis tool with a resolution ability of zero.22
supply of economy for the country. Soil is a vital key issue nm/pixel by utilizing 3D printing technology, regular
for agriculture[1]. There are many soil varieties in Asian optical parts, a Digital Versatile Disc as a grating part,and
nation. so as to predict the kind of crop which will be the rear camera of the phone. The detector responses for
cultivated in this explicit soil kind we'd like to grasp the normal pH scale samples within the pH scale vary four to
options and characteristics of the soil type. to explain and ten ar ascertained to be linear however yield a sensitivity of
classify soils giving uniform system of classification with 0.129 per pH scale unit. The resolution of the projected
uniform language so as to correlate the soils of various detector for the thought of samples is ascertained to be
space. to point out distribution of various soils within the zero.09 pH scale units. The results obtained from the
field(soil mapping).To provide knowledge for creating designed tool whereas mensuration the pH scale values of
interpretations on the ability of explicit soils for agricultural six field-collected soil samples are found to be correct. The
purpose and conjointly for several different functions (as in designed sensor’s performance has been evaluated by
soil management).For proper crop yield, farmers ought to scrutiny the experimental knowledge with the commercial-
bear in mind of the right soil kind for a selected crop, that grade pH scale sensing too.
affects the accumulated demand for food. There are
numerous laboratory and field strategies to classify soil, In the Paper ” Temporal-Spatial Soil Moisture
however these have limitations like time and labor- Estimation from CYGNSS Using Machine Learning
consuming. there's a demand of computer-based soil Regression With a Pre classification Approach ” [4] -
classification techniques which is able to facilitate farmers Global navigation satellite system-reflectometry (GNSS-R)
within the field and won’t take tons of your time. will retrieve Earth’s surface parameters, like soil wet (SM)
victimization the mirrored signals from GNSS
II. RELATED WORKS constellations with benefits of noncontact, unrestricted,
realtime, and continuity, significantly the space-borne
In the Paper [1]“ Prediction Of Land Suitability For
cyclone GNSS (CYGNSS) mission. However, the accuracy
Crop Cultivation Based On Soil And Environmental
and potency of SM estimation from CYGNSS still have to
Characteristics Using Modified Recursive Feature
be compelled to improve. during this article, the world SM
Elimination Technique With Various Classifiers “ [2] -
is calculable victimization machine learning (ML)
Crop cultivation prediction is associate integral a part of
regression power-assisted by a preclassification strategy.
agriculture and is based on factors like soil, environmental
options like precipitation and temperature, and therefore the the full observations are classified by land sorts and
quantum of fertilizer used, significantly chemical element corresponding subsets are engineered for constructing mil
and phosphorus. These factors, however, vary from region regression submodels. Ten-fold cross-validation technique
to region: consequently, farmers area unit unable to is adopted. the general performance of SM estimation
with/without preclassification is compared, and therefore
cultivate similar crops in each region. this is often wherever
the results show that the SM estimations victimization
machine learning techniques step in to assist realize the
totally different mil algorithms all have substantial
foremost appropriate crops for a selected region, so helping
improvement with the preclassification strategy. Then, the
farmers a good deal in crop prediction. The feature choice
best XGBoost foreseen model with root-mean-square error
side of metric capacity unit may be a major element within
(RMSE) of zero.052 cm3/cm3 is adopted. additionally, the

IJISRT22JUN004 www.ijisrt.com 80
Volume 7, Issue 6, June – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
satisfactory daily Associate in Nursingd seasonal SM III. PROPOSED SYSTEM
prediction outcomes with an overall correlation worth of
zero.86 Associate in Nursingd an RMSE worth of zero.056 Soils area unit of assorted sorts and each soil kind will
cm3/cm3 ar achieved at a worldwide scale, severally. have totally different composition of minerals, humus,
moreover, the in depth temporal and abstraction variations organic matter and may hold completely different
of CYGNSS SM predictions ar evaluated. characteristics supported that different crops area unit
typically full-grown. an accurate soil take a look at can
In the Paper “Agricultural Soil Data Analysis Using facilitate certify the applying of enough plant food to
Spatial Clustering Data Mining Techniques ” [5] – Soil satisfy the requirements of the crop whereas taking
could be a crucial dynamic resource that supports advantage of the nutrients already gift at intervals the soil.
vegetation. Soil data analysis can provide knowledge that is Taking soil samples, Laboratory analysis of samples is
important for increasing nutrient use efficiency and incredibly long method for farmers. therefore we have got
agricultural productivity. it is a helpful gizmo for soil creating a system for soil analysis. The model has been
management and would possibly facilitate to cut back the tested by applying totally different forms of machine
onsite and offsite costs of erosion, improve water and learning algorithms. SVM has given the best accuracy in
nutrient use efficiencies, and confirm that the resource is soil classification. The planned model is even by a properly
sustained for future use. As degree unattended learning created dataset and machine learning algorithms. To
technique, spatial agglomeration has emerged to be one in estimate the wet content within the soil is one in all the
each of the foremost important techniques at intervals the foremost tough task for humans and robots, in packaging or
sphere of agriculture for soil data analysis. Soil properties segregation or choose and place operations.
(including physical, chemical, and biological properties)
and thus the characteristics of the spatial soil data area unit Our objective is to help the farmers to know about
first introduced. spatial agglomeration techniques area unit fertility of the soil without testing the soil in laboratories.
then summarized in five fully completely different They can itself test the soil with the help of a smartphone .
categories.Soil data analysis practice spatial agglomeration By just simply taking the picture of the soil they can get
is reviewed in four categories of agricultural applications: soil testing results in less time. From these results the
agricultural production management partitioning, farmer can know the fertility of the soil , pH value, and it
comprehensive assessment of soil and land, soil and land will also recommend the crops which will be suitable in
classification, and correlation study for agro-ecosystem. the that soil. It will recommend what are the fertilizers are need
quality agglomeration algorithms usually work well, and to be used for the particular crop if we are cultivating in
prototype-based agglomeration ways that area unit further that soil. It shows the amount of fertilizers, pH value and
preferred in observe. Some machine learning models is any other contents in the soil. It will recommend the crops
introduced into the spatial agglomeration algorithms for which will give good yield in that soil and also recommend
higher accommodation to various characteristics of soil which all fertilizers are to be used for the soil. The
dataset. proposed work is to develop a Machine learning based soil
classification and crop recommendation system. A large
In the paper “ Comparative Analysis of Soil Properties data set of collected images in various categories is used to
to Predict Fertility and Crop Yield using Machine Learning train the system. Collected data set is divided into training,
Algorithms “[6] - Agriculture is a vital a part of human validation and testing set for training and examining the
lives. it's one in every of the most important supply of framework. In the proposed system the images are acquired
employment in India. More than 1/2 the population depend from the user. The images are received from the farmer via
on agriculture. it's the backbone of our economy. Crop the Android Application developed exclusively for the
yield depends on several factors. one in every of the most service of the farmer and common men. The images are
important factors that have an effect on the yield of the crop uploaded by the user by choosing the appropriate image of
is soil. Improvising the techniques to predict crop yield in the soil or the piece of land picture from the Choose File
several atmospheric condition will facilitate farmers and option. The image uploaded by the farmer is processed by
other stakeholders in higher higher cognitive process in the model. Then image processing techniques are applied to
terms of scientific agriculture and crop choice. Crop yield extract useful features from the images, that are necessary
prediction includes foretelling the yield of the crop from for further analysis.
previous historical knowledge that consists of things like
temperature, humidity, pH, rain and crop name. It provides IV. METHODOLOGY
United States of America an inspiration for the best
foreseen crop which can be cultivate within the field Soil Image Dataset and crop which is the main task of
weather.In the planned work, a comparative analysis on soil the project to collect the soil images .The dataset is
properties to predict fertility and crop yield has been consisting of different soil types and their features and
performed mistreatment machine learning algorithms. The different types of crops suitable for the soil. On this soil
analysis has been done on self –obtained dataset, for 3 data we are training and testing the model for soil
crops - tomato, potato and chili. The crop yield prediction classification and crop Recommendation system. After
has been done mistreatment K Nearest Neighbour collection the soil an crop data the next step is
algorithmic program, Naïve Thomas Bayes algorithmic preprocessing. We can see it in the architecture [Fig.1].
program and call Trees classifier. Preprocessing, its main aim is to improve the quality of the
images of the soils which we are taken using camera’s etc.

IJISRT22JUN004 www.ijisrt.com 81
Volume 7, Issue 6, June – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
It is having different steps inside the preprocessing to
format the image. Color correction, resizing etc . Then Data
is spliting, To soil data and crop data. The soil data consist
of the soil types and its properties, the crop data consist of
which of the grow in that particular soil. Dataset Deviation
Here the existing Data is taken for both Training and
Testing of the data.

A. Architecture
In an large data set more number of data consisting of
soil images and their features will undergo for Training the
model. After the Training of the model only by using our
we can predict the soil type and its features. And the
remaining data will be used for testing the model which
trained. This Data set of soil which will be divided for both
the training and testing. In an large data set more number of
data consisting of soil images and their features will Fig. 2: Flowchart of the application
undergo for Training the model. After the Training of the
model only by using our we can predict the soil type and V. RESULT
its features. And the remaining data will be used for testing Our product can give our farmers a better knowledge
the model which trained. The crop data set which will be
about the suitable crop for the land with ease of using of the
also divided to both training and testing. In an large data set product. It also gives better accuracy than existing systems.
more number of data consisting of crop images and their
When we compared this product with another product it is
details will undergo for Training the model. After the proved that pH, nutrients predicted by the prodcut is more
Training of the model only by using our we can predict the accurate than the that of another product. The temperature
crop which can be cultivated in that soil. And the remaining and humidity shown by the this product is alsmost same as
data will be used for testing the model which trained. the orginal temperature and humidity.

VI. CONCLUSION

Agriculture is that the major source for living for the

people of India. Agriculture research is that the major
source of economy for the country. To supply data for
creating interpretations as to the adaptability of particular
soils for agricultural purpose. For correct crop yield,
farmers should bear in mind of the proper soil type for a
specific crop, which affects the increased demand for food.
There are different laboratory and field methods to classify
soil, but they also have limitations. So There's a
requirement of computer-based soil classification
techniques which can help farmers and won’t take plenty
of time.

REFERENCES

[1.] Laishram, Joyalata Saxena, Krishna Maikhuri, Rakesh

Rao, Kottapalli. (2012). Soil quality and soil health: A
Fig. 1: Architecture of the application review.. International Journal of Ecology and
Environmental Sciences. 38.
B. Flowchart [2.] G. Mariammal , A. Suruliandi , S. P. Raja , and E.
The flowchart [Fig.2] which is able to show the flow of Poongothai, “Prediction of Land Suitability for Crop
the method of our system. First it opens the applying and so Cultivation Based on Soil and Environmental
image is taken and uploaded if the image isn't getting the Characteristics Using Modified Recursive Feature
standard the image are going to be retaken and again Elimination Technique With Various Classifiers ”,
uploaded and processed. The soil image is processed so the IEEE TRANSACTIONS ON COMPUTATIONAL
soil type is predicted if not it'll be asked to retake the image SOCIAL SYSTEMS, VOL. 8, NO. 5, OCTOBER
of the soil. After predicting what form of soil it'll also 2021.
predicts its features which can be usefull for the crop [3.] Priyanka Das, Sarmistha Paul, Satya Sundar
recommendation Then it'll predict the crop which can be Bhattacharya, and Pabitra Nath ,” Smartphone-Based
suitable for that soil. Here both the predictions are done by Spectrometric Analyzer for Accurate Estimation of
different models. this can be the most flow of the system pH Value in Soil” ‘,IEEE SENSORS JOURNAL,
we had proposed. VOL. 21, NO. 3, FEBRUARY 1, 2021.

IJISRT22JUN004 www.ijisrt.com 82
Volume 7, Issue 6, June – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[4.] Shuanggen Jin , Haolin Chen, Qingyun Yan , Patrizia
Savi , Yan Jin , Yan Jia, and Yuan Yuan , “Temporal-
Spatial Soil Moisture Estimation from CYGNSS
Using Machine Learning RegressionWith a
Preclassification Approach” ,IEEE. IEEE JOURNAL
OF SELECTED TOPICS IN APPLIED EARTH
OBSERVATIONS AND REMOTE SENSING, VOL.
14, 2021.
[5.] Hongju Gao ,” Agricultural Soil Data Analysis Using
Spatial Clustering Data Mining Techniques ”, 2021
IEEE 13th International Conference on Computer
Research and Development.
[6.] Pranay Malik, Sushmita Sengupta, Jitendra Singh
Jadon, ” Comparative Analysis of Soil Properties.
[7.] Predict Fertility and Crop Yield using Machine
Learning Algorithms “ , 2021 11th International
Conference on Cloud Computing, Data Science &
Engineering (Confluence).
[8.] Jie Ren , Jing Liang , and Yuanyuan Zhao “ Soil PH
Measurement Based on Compressive Sensing and
Deep Image Prior “ , IEEE TRANSACTIONS ON
EMERGING TOPICS IN COMPUTATIONAL
INTELLIGENCE, VOL. 4, NO. 1, FEBRUARY
2020.
[9.] Vitalik Buterin, Evgeniou, Theodoros Pontil,
Massimiliano. (2001). Support Vector Machines:
Theory and Applications. 2049. 249-257. 10.1007/3-
540-44673-712.
[10.] M. Gopal P S and B. R, “Selection of important
features for optimizing cropyield prediction,” Int. J.
Agricult. Environ. Inf. Syst., vol. 10, no. 3, pp. 54–
71,Jul. 2019.
[11.] F. Balducci, D. Impedovo, and G. Pirlo, “Machine
learning applications on agricultural datasets for smart
farm enhancement,” Machine, vol. 6, no. 3, pp. 38–59,
2018.

IJISRT22JUN004 www.ijisrt.com 83