KRISHI SAMRIDHI

Submitted in partial fulfillment of the requirements

of the degree of

Bachelor of Engineering
by

Anujeet Kunturkar (30)

Mervis Mascarenhas (34)
Nilesh Mishra (35)
Nadar Maheshwaran Ganeshan (37)

Supervisor:
Prof. Teena Varma

Computer Engineering Department

Xavier Institute of Engineering
University Of Mumbai
2020-2021
 CERTIFICATE

This is to certify that the project Synopsis entitled “Krishi Samridhi” is a bonafide work
of

Anujeet Kunturkar (30)

Mervis Mascarenhas (34)
Nilesh Mishra (35)
Nadar Maheshwaran Ganeshan (37)

submitted to the University of Mumbai in partial fulfillment of the requirement for the award
of the degree of “Bachelor of Engineering” in “Computer Engineering”.

images/Internal Examiner.png
————————————–
(Prof. Teena Varma)

Supervisor/Guide

Prof. Kavita Dr. Y.D. Venkatesh

Head of Department of Computer Engineering Principal
 Project Report Approval for B.E.

This project entitled “Krishi Samridhi” by

Anujeet Kunturkar (30)

Mervis Mascarenhas (34)
Nilesh Mishra (35)
Nadar Maheshwaran Ganeshan (37)

is approved for “Bachelor of Engineering” in “Computer Engineering” by University

of Mumbai during the academic year 2020-2021.

Examiner:

images/External_Examiner.pn

1. Dr. Nilesh M. Patil

(External Examiner)

images/Internal Examiner.p

2. Teena Varama
(Internal Examiner)

Date:
Place:
 Declaration

I declare that this written submission represents my ideas in my own words and where others’
ideas or words have been included, I have adequately cited and referenced the original sources.
I also declare that I have adhered to all principles of academic honesty and integrity and have
not misrepresented or fabricated or falsified any idea/data/fact/source in my submission. I
understand that any violation of the above will be cause for disciplinary action by the Institute
and can also evoke penal action from the sources which have thus not been properly cited or
from whom proper permission has not been taken when needed.

images/Sign.jpg
———————————————
(Anujeet Kunturkar (30))

images/Sign.jpg
———————————————
(Mervis Mascarenhas (34))

images/Sign.jpg
———————————————
(Nilesh Mishra (35))

images/Sign.jpg
———————————————
(Maheshwaran Nadar (37))

Date:
 Abstract

The project ”Krishi-Samridhi” aims to revolutionize the agricultural sector by integrating

cutting-edge technology to provide crucial insights into crop health and productivity. The
solution combines plant leaf disease detection, crop recommendation, and soil moisture
measurement into a unified platform. Leveraging IoT devices, machine learning (ML) models,
and cloud-based infrastructure (Firebase), this system enables real-time monitoring and
data-driven recommendations to help farmers make informed decisions. This project also
features a website built using ReactJS and Firebase for intuitive user interaction, which
integrates the machine learning models and displays results based on the inputs from the IoT
devices.
 Contents

1 Introduction 9
1.1 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Aims and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.1 Aim: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.2 Objectives: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Scope of the Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Review of Literature 11
2.1 Survey of Existing Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Limitations of Existing Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Literature Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Description 13
3.1 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1 Use Case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Feasibility Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.2 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.3 Analysis and Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.4 Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Implementation Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3.1 Problem Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3.2 Data Collection and Preprocessing . . . . . . . . . . . . . . . . . . . . . 18
3.3.3 Leaf Disease Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3.4 Crop Recommendation System . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Details of Hardware and Software . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.1 Firebase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.2 VGG19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.3 YOLOv8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.4 Support Vector Machine (SVM) . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.5 Random Forest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4.6 Logistic Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4.7 Naive Bayes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

5
 3.4.8 Decision Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4.9 NPK Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4.10 Temperature and Humidity Sensor . . . . . . . . . . . . . . . . . . . . . 21
3.4.11 Soil Moisture Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4 Implementation Plan for Next Semester 22

4.1 Backend Development and Integration: . . . . . . . . . . . . . . . . . . . . . . . 22
4.2 Testing and Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Intrusion Detection System: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

5 Conclusions 24

6 Acknowledgement 25

6
 List of Figures

3.1 Use Case Diagram of our Proposed System . . . . . . . . . . . . . . . . . . . . . 13

3.2 Flowchart explaining the algorithm for training our model . . . . . . . . . . . . 17

7
 Notation and Nomenclature

D Discriminator Network

G Generator Network

8
 Chapter 1
Introduction

Agriculture plays a pivotal role in feeding the global population, particularly in rural areas
where farming is a primary source of livelihood. However, farmers face significant challenges,
such as the inability to detect crop diseases early, poor soil health management, and the lack of
reliable data-driven crop recommendations. These factors severely impact crop yield and overall
productivity, resulting in economic losses. To address these issues, the Krishi Samriddhi Project
introduces a comprehensive technological solution that integrates IoT-based soil monitoring,
leaf disease detection, and machine learning-driven crop recommendations, providing farmers
with actionable insights to improve agricultural output.

1.1 Problem Definition

Farmers encounter multiple challenges that hamper their ability to optimize crop production.
One major issue is the timely identification of leaf diseases, which is often missed until it’s too
late to prevent significant crop loss.
Additionally, maintaining proper soil health is a complex task, with farmers struggling to
monitor soil nutrients, moisture, and other key factors in real-time.
Furthermore, selecting the right crop to plant based on the current environmental and soil
conditions is another challenge, often leading to suboptimal yields. Currently available tools
either fail to offer real-time insights or are not accessible to small-scale farmers due to high
costs and complexity.

1.2 Aims and Objectives

1.2.1 Aim:
The primary aim of the Krishi Samriddhi Project is to develop an integrated system that
aids farmers in improving crop yield by providing real-time leaf disease detection, soil health
monitoring, and crop recommendations. Using advanced technologies like IoT sensors, deep
learning, and machine learning, the system will offer timely insights to enhance decision-making
and optimize farming practices.

9
1.2.2 Objectives:
• Implement a deep learning model (VGG19) for early classification of leaf diseases.

• Use YOLOv8 to detect and localize diseased areas in real time.

• Deploy IoT sensors to continuously monitor soil conditions, including NPK levels, mois-
ture, temperature, and humidity.

• Perform a comparative study of machine learning models to recommend the best crop,
with Random Forest used for final recommendations based on its high accuracy.

• Provide farmers with real-time data and actionable insights through a user-friendly plat-
form for better crop management decisions.

1.3 Scope of the Project

The scope of this project involves developing an integrated system that addresses multiple
aspects of crop management. The project will leverage IoT sensors to collect real-time data on
soil nutrients, moisture, temperature, and humidity, storing this data in a cloud-based database
(Firebase). Simultaneously, high-resolution images of leaves will be analyzed using the VGG19
deep learning model to classify diseases, and YOLOv8 will be employed to pinpoint the infected
regions. On the crop recommendation side, the system will use a combination of machine
learning models, ultimately selecting Random Forest for its superior accuracy in predicting
suitable crops based on real-time sensor data. The project is designed to be scalable and
adaptable, allowing the integration of additional crops, soil types, and environmental factors in
the future.

10
 Chapter 2
Review of Literature

2.1 Survey of Existing Systems

Several technologies exist for monitoring various aspects of agriculture. Current systems focus
either on soil moisture, disease detection, or yield optimization individually. IoT-based solutions
have been developed to measure soil moisture, but these systems often lack integration with
ML models to offer intelligent recommendations. Similarly, plant disease detection has been
explored using image classification techniques but without comprehensive data pipelines that
combine real-time IoT data and cloud-based analysis. These systems often provide fragmented
solutions rather than a unified platform for end-to-end farm management.

2.2 Limitations of Existing Systems

• Lack of Integrated Solutions: Most existing systems provide isolated solutions for specific
agricultural problems such as disease detection or soil moisture monitoring but lack a
comprehensive, unified platform.

• Limited Data Utilization: Current systems do not fully leverage IoT data in combination
with ML models for predictive analytics or crop recommendations.

• Accessibility: The usability of many platforms is constrained by technological complexity,

lacking a user-friendly interface tailored for farmers.

• Delayed Response Time: Existing systems often fail to deliver real-time recommendations,
which can lead to reduced farm productivity.

11
2.3 Literature Table

Title Model Advantage Disadvantage

Plant Leaf Disease De- Support Vector Benefit farmers in The application of
tection, Classification, Machine (SVM), achieving expedi- these techniques in
and Diagnosis Using K-Nearest Neighbors tious and appropriate the recognition of
Computer Vision and (KNN), Random actions to avoid a re- disease can avert the
Artificial Intelligence: Forest, Deep Learning duction in the quality disadvantage of origin
A Review Models: 1. Con- and quantity of crops. by a factious selection
volutional Neural of disease features,
Networks (CNNs), extraction of features,
2. Recurrent Neural and boost the speed
Networks (RNNs) of technology and
efficiency of research.
Plant Disease Detec- Support Vector Ma- Multiple diseases on The model has poor
tion and Classification chines (SVM), CNN, the same leaf could be robustness.
by Deep Learning VGG16, GANs, detected, and the data
ResNet can be augmented by
cutting up the leaf im-
age into multiple sub-
images.
Plant Disease Detec- 1. Logistic regres- 1. Random Forests 1. Requires more
tion Using Machine sion: 65.33, 2. Sup- overcome the disad- computational re-
Learning port vector machine: vantage of overfitting sources for training
40.33, 3. K-nearest typically seen with de- and prediction com-
neighbor: 66.76, 4. cision trees by averag- pared to simpler
CART: 64.66, 5. Ran- ing multiple trees. 2. models. 2. Needs
dom Forests: 70.14, 6. Random Forests can more memory to store
Naı̈ve Bayes: 57.61 be used for both clas- the multiple trees
sification and regres- used in the model.
sion tasks.
Plant Disease De- 1. Custom CNN, The small transformer Proper application
tection Using Deep 2. INCEPTIONv3, 3. network (STN) seems of transformers in
Learning STN, 4. LTN to be the best fit for computer vision tasks
applications requiring needs more research
minimal resources. as this architecture is
in its early days.
Leaf Disease Detec- K-means clustering, 1. Helps in identify- 1. Requires the num-
tion and Classifica- GLCM for Feature ing disease-affected ber of clusters (K)
tion based on Machine Extraction, SVM for regions effectively. to be specified in ad-
Learning Classification 2. Efficient for large vance. 2. Computa-
datasets. tionally intensive, es-
pecially for large im-
ages.
12
 Chapter 3
Description

3.1 Analysis
In the previous sections, we have defined our problems clearly and surveyed the existing liter-
ature and approaches. Now we discuss how we analyzed this problem from our prospective.

3.1.1 Use Case Diagram

Figure 3.1: Use Case Diagram of our Proposed System

13
3.1.2 Feasibility Study
Here we discuss if we can achieve the planned aims of our project while staying within the
constraints of our resources.

• Cost of data acquisition

– How hard is it to acquire data?

– How expensive is data labelling
– How much data will be needed

• Cost of wrong Prediction

– How frequently does the system need to be right to be useful?

• Availability of good published work about similar problems

– Has the problem been reduced to practise

– Is there sufficient Literature on the problem

• Computational Resources available for both training and inference

– Will the model be deployed in a resource constrained environment

3.1.2.1 Cost of Data Acquisition

• Plant leaf disease detection, crop recommendation, and soil moisture monitoring require
a substantial amount of data from agricultural sources. While there are some available
datasets for crop recommendations and soil health, curating a plant disease dataset spe-
cific to regional crops may require collaboration with agricultural experts or field data
collection.

• Data labeling can be costly and time-consuming, especially for plant disease detection
where expert knowledge is required to correctly label various diseases. We might need to
rely on agricultural scientists or AI-based pre-labeling methods, which can also require
manual verification.

• For the crop recommendation system and soil moisture predictions, moderate-sized datasets
can provide good results if the model is well-tuned. For plant disease detection, we re-
quire high-quality images and labeled data to improve accuracy, as the problem is more
complex.

14
3.1.2.2 Cost of wrong Prediction
• For plant disease detection, a wrong prediction can lead to incorrect treatments, poten-
tially harming crops. The system should aim for at least 8/10 correct predictions to
ensure that farmers can trust its recommendations.

• For crop recommendation, incorrect predictions could lead to poor yields, and the system
should also aim for 7-8/10 accuracy to be effective.

• For soil moisture monitoring, the impact of incorrect predictions is lower, but incorrect
readings can lead to inefficient irrigation, so a 6/10 accuracy may suffice in some cases.

3.1.2.3 Availability of good published work about similar problems

• Several research papers exist on crop recommendations, soil health monitoring, and dis-
ease detection, but many of these are in controlled environments or specific regions. The
challenge lies in adapting those models to a wider range of environmental factors.

• While crop recommendation and soil moisture analysis are well-researched areas, plant
disease detection using machine learning, especially tailored for specific regional crops,
still lacks extensive literature. However, the research community is growing rapidly in
this area, providing a foundation for further work.

3.1.2.4 Computational Resources available for both training and inference

• Training: The training process for plant disease detection models, especially if using deep
learning, requires substantial computational resources (GPUs). For crop recommendation
and soil moisture, simpler models (e.g., decision trees, linear regression) can be trained
on smaller datasets and with lower hardware requirements.

• Inference: Once deployed, the models should be optimized to run on low-resource devices.
The aim is to make the Krishi-Samriddhi platform accessible to farmers using smartphones
or low-cost computing devices.

15
3.2 Design
The design of the Krishi Samriddhi Project integrates advanced hardware and software com-
ponents to provide a comprehensive solution for sustainable farming. The system is designed
to monitor environmental conditions, detect plant diseases, and recommend suitable crops for
optimal yield. This section outlines the key design elements that form the backbone of the
project.

3.2.1 Data Collection

The system uses IoT-based sensors for real-time data collection from the field. These sensors
continuously monitor the soil and environmental conditions, such as soil nutrient levels (NPK),
moisture, temperature, and humidity. The collected data is sent to a central system through a
microcontroller for further processing and analysis.

3.2.2 Data Processing

The sensor data is stored in Firebase, a cloud-based database that ensures real-time data access.
This centralized storage allows the system to access and analyze the data efficiently. The data
from the sensors undergoes cleaning and preprocessing to remove any noise or inconsistencies
before further analysis.

3.2.3 Analysis and Output

• Leaf Disease Detection: The leaf images captured by the system are processed using
the VGG19 convolutional neural network (CNN) model for classification. This model is
trained on a large dataset of diseased and healthy leaf images to identify the presence of
diseases.

• YOLOv8 Object Detection: The diseased areas on the leaves are detected using the
YOLOv8 model. It performs real-time object detection, allowing the system to not only
classify the disease but also localize the infected regions.

• Crop Recommendation System: Based on the data collected from the sensors (such
as NPK levels, soil moisture, temperature, and humidity), the system uses a compara-
tive study of machine learning models to recommend the most suitable crops. Random
Forest, which provided the highest accuracy in the study, is used to make the final crop
recommendations.

16
3.2.4 Flowchart
In this section, we explain the high-level steps involved during the various phases of our project
using flowcharts.

3.2.4.1 Flowchart detailing the training process

Figure 3.2: Flowchart explaining the algorithm for training our model

17
3.3 Implementation Methodology
The Krishi Samriddhi Project aims to tackle the challenges faced by farmers due to crop
diseases, unpredictable weather, and inadequate crop selection by utilizing IoT-based sensors
and machine learning models. Our approach integrates advanced technologies to provide real-
time monitoring, disease detection, and crop recommendations, which empower farmers with
data-driven insights to optimize their agricultural practices.

3.3.1 Problem Identification

The project addresses three key challenges in modern agriculture: early detection of leaf dis-
eases, real-time monitoring of soil and environmental conditions, and optimized crop recom-
mendation. The early detection of leaf diseases is crucial to prevent crop loss, ensuring timely
intervention before the disease spreads. By employing advanced machine learning techniques,
we can analyze images of leaves and detect infections at an early stage. Additionally, the
project monitors soil moisture, temperature, humidity, and nutrient levels through sensors to
optimize irrigation and nutrient supply, ensuring that crops receive the right conditions for
growth. Furthermore, our crop recommendation system, based on soil characteristics, weather
patterns, and historical yield data, suggests the most suitable crops to maximize productivity,
helping farmers make informed decisions for improved yield and sustainability.

3.3.2 Data Collection and Preprocessing

• Hardware-based Data Collection We deployed various IoT sensors (NPK sensor,
temperature and humidity sensor, and soil moisture sensor) in the field to gather real-
time data on the key parameters affecting crop health and growth. These sensors provide
continuous updates on soil nutrient levels, moisture content, temperature, and humidity,
which are crucial inputs for both disease detection and crop recommendation models.

• Image-based Data Collection We captured images of leaves using high-resolution

cameras for disease detection. These images are used as input to the machine learning
models for classification and object detection.

• Data Storage All sensor data and images were stored in a centralized Firebase database,
ensuring scalable and real-time data access for further analysis.

• Preprocessing Sensor data was preprocessed to handle missing or inconsistent values,

normalize readings, and filter out noise. Image data underwent augmentation techniques
such as rotation, scaling, and cropping to improve model generalization and ensure ro-
bustness in disease detection.

3.3.3 Leaf Disease Detection

To address the problem of early disease detection in crops, we applied a deep learning-based
approach using the VGG19 Convolutional Neural Network (CNN) model:

18
 • Trainin: The model was trained on a large dataset consisting of both healthy and diseased
leaf images.

• Classification: The model predicts whether a leaf is healthy or infected based on the
input images.

• YOLOv8 Object Detection: The object detection model was further applied to localize
the diseased areas in the infected leaves. YOLOv8, due to its fast real-time detection
capabilities, allowed us to not only classify the disease but also pinpoint the exact areas
affected, helping farmers apply targeted interventions.

3.3.4 Crop Recommendation System

For crop recommendation, we analyzed the sensor data to provide farmers with customized
suggestions on which crops to grow. Our methodology consisted of:

• Data Analysis: We collected environmental data (NPK levels, soil moisture, tempera-
ture, and humidity) from the sensors.

• Model Training: Five machine learning models — Support Vector Machine (SVM),
Random Forest, Logistic Regression, Naive Bayes, and Decision Tree — were trained on
historical crop yield data and corresponding environmental conditions.

• Comparative Study: Each model’s performance was evaluated based on accuracy,

precision, and recall. Random Forest outperformed the others, achieving the highest
accuracy, and was therefore selected as the final model for crop recommendations.

• Prediction: The Random Forest model analyzed the sensor data and predicted the most
suitable crops for a given field, taking into account the soil health and environmental
conditions.

19
3.4 Details of Hardware and Software
The hardware with which we initially started was an AMD Radeon R5 M330 2GB GPU, but
as the training of our model was very resource heavy, it was putting a strain on our hardware,
hence we decided to use the GPU provided by Google Colaboratory, along with our hardware
to reduce our training time and not put too much load on our hardware.

3.4.1 Firebase
All sensor data, including NPK values, temperature, humidity, and soil moisture levels, are
stored in Firebase. This cloud platform provides real-time data access, enabling farmers to
analyze their data and make decisions accordingly.

3.4.2 VGG19
VGG19 is a deep learning model used to classify leaf diseases. By training the model on large
datasets of leaf images, the system can accurately identify diseases in crops. Early detection
enables quick intervention, preventing significant damage and loss.

3.4.3 YOLOv8
YOLOv8 performs object detection on leaf images, identifying areas affected by disease. It
provides real-time detection with high precision, allowing farmers to detect issues early and
treat them before they spread, thereby reducing crop damage.

3.4.4 Support Vector Machine (SVM)

SVM is a supervised learning model used for classification and regression analysis. In the crop
recommendation system, SVM helps in classifying different types of crops based on features
like soil type, nutrient levels, and environmental conditions. Though it performed well, SVM
didn’t provide the highest accuracy when compared to other models in this project.

3.4.5 Random Forest

Random Forest is an ensemble learning method used for classification and regression. It builds
multiple decision trees and merges them to get a more accurate prediction. In this project,
Random Forest outperformed the other models, giving the best accuracy for crop recommen-
dation. By analyzing various factors such as soil health, climate, and historical crop data, it
provides precise suggestions on which crops to plant.

20
3.4.6 Logistic Regression
Logistic Regression is a statistical model that uses a logistic function to model binary-dependent
variables. In the context of crop recommendation, it helps in predicting whether a specific crop
is suitable for the given environmental and soil conditions. While it provides good results for
binary decisions, its performance was lower compared to Random Forest in this case.

3.4.7 Naive Bayes

Naive Bayes is a probabilistic machine learning model based on Bayes’ Theorem. It works
well for classification tasks with independent features. For crop recommendation, Naive Bayes
classifies crop suitability based on individual factors like soil type, climate, and moisture levels.
However, its simplicity led to less accurate results than the more complex Random Forest
model.

3.4.8 Decision Tree

Decision Tree is a model that makes decisions based on the values of various features. It
creates a tree structure where each internal node represents a feature, each branch represents a
decision rule, and each leaf node represents an outcome. In the crop recommendation system,
the Decision Tree model was used to predict the best crop based on soil and environmental
data. Although it performed well, Random Forest, which aggregates multiple decision trees,
provided more accurate results.

3.4.9 NPK Sensor

This sensor measures the concentration of nitrogen, phosphorus, and potassium in the soil.
These nutrients are crucial for plant growth, and real-time monitoring allows farmers to make
informed decisions about fertilizer use, improving crop health and yield.

3.4.10 Temperature and Humidity Sensor

Monitoring environmental conditions is essential for optimal crop growth. The temperature
and humidity sensor ensures that crops are growing in the ideal climate, allowing farmers to
adapt to changes in weather and avoid any adverse effects.

3.4.11 Soil Moisture Sensor

By measuring the water content in the soil, this sensor ensures that crops receive the right
amount of water. This helps in preventing both over-watering and drought conditions, thereby
optimizing water usage and improving crop production efficiency.

21
 Chapter 4
Implementation Plan for Next
Semester

4.1 Backend Development and Integration:

Build a robust backend to handle real-time data, manage predictions, and interface with hard-
ware devices.

• Tasks:

– Design APIs for fetching and sending real-time data (e.g., soil moisture, tempera-
ture).
– Develop the logic to run the crop recommendation model based on real-time data.
– Secure the backend using authentication (JWT, OAuth) for admin and user access.

• Tech Stack:

– Node.js/Express for server-side logic.

– MongoDB or Firebase for database management.
– RESTful APIs for communication between frontend and backend.

4.2 Testing and Validation

Validate the system by testing it with real-time data from the sensors and improve prediction
accuracy.

• Tech Stack:

– Run multiple tests in varying conditions (different soil moisture levels, temperatures).
– Debug and optimize both hardware and software for real-world performance.
– Gather user feedback for UI/UX improvements.

22
4.3 Intrusion Detection System:
• Add an IDS to monitor and improve farm security.

• Use IoT sensors and machine learning algorithms to detect unauthorized access or unusual
activities.

• Provide real-time alerts through the existing web application for easy monitoring.

23
 Chapter 5
Conclusions

The implementation of KrishiSamridhi aims to provide a comprehensive solution for farmers,

combining modern technology with traditional agricultural practices. The successful integration
of real-time data collection, hardware sensors, and backend processing will enable precise and
timely crop recommendations, significantly enhancing decision-making for farmers.
By leveraging sensor data (soil moisture, temperature, humidity) and utilizing a crop rec-
ommendation algorithm, KrishiSamridhi offers a scalable and practical solution to improve
agricultural productivity. The integration of hardware components like Arduino or Raspberry
Pi with real-time data processing ensures that the system remains both accurate and responsive
to changing environmental conditions.
This project’s real-world application potential makes it a valuable tool for addressing chal-
lenges faced by farmers, particularly in regions where precise agricultural practices are crucial.
The next phase of the project will involve real-time integration, testing, and deployment, mak-
ing KrishiSamridhi a fully functional and impactful system. Upon completion, the system will
empower farmers with data-driven insights, contributing to sustainable and efficient farming
practices.
The journey ahead, with the focus on hardware integration and real-time processing, promises
not only technical growth but also practical solutions for agriculture, paving the way for inno-
vation in the sector.

24
 Chapter 6
Acknowledgement

We sincerely thank our mentor and guide Prof. Teena Varma for her constant guidance and
help in our research. Without her tutelage, we wouldn’t be able to complete this project. She
encouraged us to work on this project.
We are also grateful to our college for giving us this opportunity to work with them and
providing us with the necessary resources for the project. Working on this project also helped
us to do lots of research and we came to know about many new things.
We are immensely grateful to all involved in this project as without their inspiration and
valuable suggestions it would not have been possible to develop this project within the pre-
scribed time.

images/Sign.jpg
———————————————
(Anujeet Kunturkar (30))

images/Sign.jpg
———————————————
(Mervis Mascarenhas (34))

images/Sign.jpg
———————————————
(Nilesh Mishra (35))

images/Sign.jpg
———————————————
(Maheshwaran Nadar (37))

25