AgriSense AI

Integrated AI for Sustainable Farming and Enhanced

Productivity

Submitted by
Abhay Kumar (22BCE11152),
Aadish Pankaj Dabhade (22MIP10053),
Aditya Sachin Khamitkar (22MIP10006),
Angiras Devendra Bawkar (22BAS10004),
Patil Devendra Prashant (22MIP10063),
Pranav Sanjay Jahagirdar (22BCY10233),
Vaibhavi Rajguru (22BAS10006)

In Partial Fulfillment of the Requirements for the Award of the Degree of

Master of Technology
(5-Year Integrated Program)

VIT Bhopal University

Bhopal, Madhya Pradesh

December 2024
 VIT Bhopal University
Bhopal, Madhya Pradesh

Bonafide Certificate

Certified that this project report titled “AgriSense AI: Integrated AI for
Sustainable Farming and Enhanced Productivity” is the bonafide work of

Sl. No. Register Number Name

1 22BCY10233 Pranav Sanjay Jahagirdar
2 22MIP10006 Aditya Sachin Khamitkar
3 22MIP10063 Patil Devendra Prashant
4 22BCE11152 Abhay Kumar
5 22BAS10004 Angiras Devendra Bawkar
6 22MIP10053 Aadish Pankaj Dabhade
7 22BAS10006 Vaibhavi Rajguru

who carried out the project work under my supervision.

This project report (Phase I) is submitted for the Project Viva-Voce examination held
between the 9th and13th of December2024

Dr. Adarsh Patel

Supervisor

Dr. Anil Kumar Yadav Dr. Rajendra Mahanandia

Reviewer Reviewer
 VIT Bhopal University
Bhopal, Madhya Pradesh

Declaration of Originality

We, hereby declare that this report entitled “AgriSense AI: Integrated AI for
Sustainable Farming and Enhanced Productivity” represents our original work
carried out for the EPICS project as students of VIT Bhopal University. To the best of
our knowledge, it contains no material previously published or written by another
person, nor any material presented for the award of any other degree or diploma of VIT
Bhopal University or any other institution. Works of other authors cited in this report
have been duly acknowledged under the section “References.”

Date: Reg. No. and Names:

Sl. No. Register Number Name

1 22BCY10233 Pranav Sanjay Jahagirdar
2 22MIP10006 Aditya Sachin Khamitkar
3 22MIP10063 Patil Devendra Prashant
4 22BCE11152 Abhay Kumar
5 22BAS10004 Angiras Devendra Bawkar
6 22MIP10053 Aadish Pankaj Dabhade
7 22BAS10006 Vaibhavi Rajguru
Contents

Acknowledgments v

Abstract vi

1 Introduction 1
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.1 Irrigation Management . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.2 Crop Yield Prediction . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.3 Supply Chain Optimization . . . . . . . . . . . . . . . . . . . . . 2
1.3.4 Vegetation Index and Drone-Based Imaging . . . . . . . . . . . . 2
1.3.5 Accessibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.6 Irrigation Management . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.7 Crop Yield Prediction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.8 Supply Chain Optimization . . . . . . . . . . . . . . . . . . . . . 3
1.3.9 Accessibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.1 Why Sustainable Farming? . . . . . . . . . . . . . . . . . . . . . . 3
1.4.2 Role of Technology in Agriculture . . . . . . . . . . . . . . . . . . 4
1.5 Implementation Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5.1 Data Collection and Analysis . . . . . . . . . . . . . . . . . . . . 4
1.5.2 User Engagement and Training . . . . . . . . . . . . . . . . . . . 4
1.5.3 Partnerships and Collaborations . . . . . . . . . . . . . . . . . . . 5
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Objectives 6
2.1 Primary Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Secondary Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Alignment with Sustainable Development Goals . . . . . . . . . . . . . . 7
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 System Design and Architecture 8

3.1 System Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Workflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Data Pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 System Features and Scalability . . . . . . . . . . . . . . . . . . . . . . . 9
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

i
4 Project Modules and Features 11
4.1 Data Acquisition and Integration . . . . . . . . . . . . . . . . . . . . . . 11
4.2 Crop and Soil Health Monitoring . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Irrigation Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.4 Yield Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.5 Crop and Irrigation Advisory . . . . . . . . . . . . . . . . . . . . . . . . 13
4.6 Supply Chain Optimization . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.7 Tech Stack Recommendations . . . . . . . . . . . . . . . . . . . . . . . . 14
4.8 AI Algorithms and Models . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.9 Additional Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.10 Development Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.11 NDVI and Vegetation Index Module . . . . . . . . . . . . . . . . . . . . 15

5 Literature Review 17
5.1 Existing Applications of AI in Agriculture . . . . . . . . . . . . . . . . . 17
5.1.1 NDVI-Based Plant Health Distribution . . . . . . . . . . . . . . . 17
5.2 Remote Sensing and GIS in Agriculture . . . . . . . . . . . . . . . . . . . 18
5.3 Addressing Knowledge Gaps . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.4 Related Work and Recent Advances . . . . . . . . . . . . . . . . . . . . . 19

6 Future Work 21
6.1 Phase 2 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.2 Evaluation and Impact Assessment . . . . . . . . . . . . . . . . . . . . . 23

7 Conclusion 24

Bibliography 25

Biodata 27

ii
List of Figures

3.1 System Architecture for AgriSense AI Platform . . . . . . . . . . . . . . 9

3.2 Data Flow Pipeline in AgriSense AI . . . . . . . . . . . . . . . . . . . . . 9

4.1 NDVI Value Ranges and Corresponding Health Classifications . . . . . . 12

4.2 KMZ Generation Workflow in AgriSense AI . . . . . . . . . . . . . . . . 12
4.3 Simulated KMZ NDVI Overlay for UAV-captured Field . . . . . . . . . . 12

5.1 NDVI-Based Plant Health Classification Distribution . . . . . . . . . . . 17

5.2 Comparison of Vegetation Health Classes Between Plot A and Plot B . . 18

6.1 Phase II Roadmap for AgriSense AI . . . . . . . . . . . . . . . . . . . . . 21

6.2 Simulated NDVI Trendlines for Two Field Plots . . . . . . . . . . . . . . 22
6.3 NDVI Evolution Over Time – Flight 1 vs Flight 2 . . . . . . . . . . . . . 22

iii
List of Tables

iv
 Acknowledgments
We express our sincere gratitude to our project supervisor, Dr. Adarsh Patel, for his
unwavering support, insightful guidance, and encouragement throughout the course of
this project. His expertise and constructive feedback have been invaluable in shaping our
work and ensuring its success.
We would also like to extend our heartfelt thanks to our reviewers, Dr. Anil Kumar
Yadav and Dr. Rajendra Mahanandia, for their critical insights and suggestions that
have greatly enhanced the quality of our project. Their thorough reviews and constructive
criticism have helped us refine our approach and focus on the key aspects of our research.
Additionally, we are grateful to the faculty members of the School of Comput-
ing and Artificial Intelligence (SCAI) for providing us with the necessary resources,
knowledge, and support throughout this project. Their dedication to fostering a con-
ducive learning environment has been instrumental in our academic journey.
We would also like to thank the School of Computing Science and Engineering
(SCOPE) and the Aerospace Department of the School of Mechanical En-
gineering (SMEC) for their collaborative spirit and support. The interdisciplinary
approach has enriched our project and provided us with diverse perspectives.
Furthermore, we extend our gratitude to the Program Chairs of the BCY, MIP,
BAS, and BCE programs for their guidance and encouragement. Their commitment
to student development has played a significant role in our academic success.
We also wish to acknowledge our peers and fellow students for their collaboration and
camaraderie, which made this project a more enriching experience. Finally, we would like
to thank our families for their encouragement and support during this endeavor.

v
Abstract

The AgriSense AI: Leveraging AI for Sustainable Agriculture and Enhanced

Productivity project introduces an innovative approach to modern farming by incor-
porating Artificial Intelligence (AI) and advanced data analytics. The primary goal of
the project is to equip farmers with actionable insights that can boost crop productivity,
optimize resource utilization, and promote sustainable agricultural practices.
AgriSense AI integrates real-time weather data, satellite imagery, and soil health
information to provide farmers with tailored, data-driven advisory services. Through
the application of advanced machine learning models, the platform delivers predictive
analytics for crucial farming decisions, such as crop selection, irrigation timing, and soil
management.
The project emphasizes environmental sustainability and features modules designed
to predict crop yields, monitor crop health, manage irrigation systems efficiently, and
streamline the agricultural supply chain. With user-friendly mobile and web applications,
AgriSense AI ensures that farmers of all technological backgrounds can easily access and
benefit from these advanced tools.
In summary, AgriSense AI aims to transform agriculture by enhancing productivity,
ensuring food security, and supporting eco-friendly farming practices, contributing to a
more sustainable and resilient agricultural future.

vi
Chapter 1

Introduction

1.1 Overview
Agriculture is the backbone of the global economy, providing essential resources such as
food, raw materials, and employment to billions of people. It plays a critical role in
sustaining livelihoods and supporting economic development, particularly in developing
countries. However, the agricultural sector faces numerous challenges that threaten its
sustainability and ability to meet the growing demands of a rapidly increasing global
population. Issues such as climate change, resource depletion, and inefficiencies in farming
practices are at the forefront of these challenges. Traditional farming methods often fail to
adapt to modern challenges, resulting in environmental degradation, loss of biodiversity,
and unsustainable growth.
Climate change has led to unpredictable weather patterns, affecting crop yields and
food security. Resource depletion, particularly of water and arable land, poses significant
risks to agricultural productivity. Furthermore, inefficiencies in farming practices, such
as over-reliance on chemical fertilizers and pesticides, contribute to soil degradation and
water pollution. These challenges necessitate a paradigm shift in how agriculture is
practiced, moving towards more sustainable and efficient methods.
AgriSense AI is designed to address these pressing issues by integrating cutting-edge
technologies such as artificial intelligence (AI), remote sensing, and Geographic Infor-
mation Systems (GIS). By analyzing diverse datasets—ranging from satellite imagery to
soil health metrics—the project empowers farmers with actionable insights that drive
productivity and sustainability. The goal is to create a comprehensive platform that not
only enhances agricultural practices but also promotes environmental stewardship and
resource conservation.

1.2 Problem Statement

The agricultural sector is plagued by inefficiencies in resource usage, leading to significant
consequences such as water scarcity, soil degradation, and reduced crop yields. Farmers
often lack access to precise data for informed decision-making, resulting in the overuse of
fertilizers, pesticides, and water. This not only affects the economic viability of farming
operations but also has detrimental effects on the environment.
For instance, excessive fertilizer application can lead to nutrient runoff, contaminat-
ing water bodies and causing algal blooms that disrupt aquatic ecosystems. Similarly,

1
inefficient irrigation practices contribute to water scarcity, particularly in regions already
facing water stress. The lack of accurate yield predictions can result in overproduction
or underproduction, leading to economic losses for farmers and food insecurity for com-
munities.
AgriSense AI addresses these critical issues by providing AI-driven recommendations
that optimize irrigation, monitor soil health, and predict crop yields. By integrating
satellite and IoT data, the project seeks to enhance decision-making processes, improve
productivity, and promote sustainable farming practices. The platform will enable farm-
ers to make data-driven decisions that align with environmental sustainability goals,
ultimately contributing to a more resilient agricultural sector.

1.3 Scope
The AgriSense AI project encompasses a wide range of capabilities designed to transform
traditional farming into a data-driven, precision-oriented operation. The updated scope
includes integration of AI, IoT, satellite and drone imagery, and user-centric advisories
for actionable decision-making.

1.3.1 Irrigation Management

AI-based irrigation advisories will be generated using both sensor-based moisture data
and imagery-derived stress analysis. This dual approach enhances water resource man-
agement in real time.

1.3.2 Crop Yield Prediction

Predictive models like XGBoost and RNNs will estimate yields based on soil, crop, and
weather data. The pipeline also supports vegetation index correlation to infer biomass
health from NDVI and NDRE patterns.

1.3.3 Supply Chain Optimization

The system will forecast harvest volume using geospatial yield maps and suggest stor-
age/transportation strategies. NDVI-mapped zones can trigger alerts for post-harvest
action based on stress levels.

1.3.4 Vegetation Index and Drone-Based Imaging

Multispectral drone cameras will provide high-resolution imagery for per-plant health
diagnostics. Using TIFF/JPG formats, the images will be analyzed for NDVI, GNDVI,
NDRE, OSAVI, and LCI indices. Output heatmaps and per-pixel classifications will be
exported to KMZ overlays for field use.

1.3.5 Accessibility
All results, advisories, and health maps will be available on mobile and web apps. Data
will be visualized through heatmaps, dashboards, and downloadable KMZs for users with
varying tech exposure.

2
1.3.6 Irrigation Management
The project will develop AI-based tools that provide precise water usage recommendations
tailored to specific crop needs and local conditions. By analyzing soil moisture levels,
weather forecasts, and crop water requirements, the system will help farmers reduce water
waste and improve irrigation efficiency. This is particularly crucial in regions facing water
scarcity, where every drop counts.

1.3.7 Crop Yield Prediction

Utilizing predictive analytics, AgriSense AI will leverage satellite data and historical
records to forecast crop yields accurately. By considering factors such as weather pat-
terns, soil health, and crop variety, the system will provide farmers with insights that
enable them to plan their planting and harvesting schedules effectively. Accurate yield
predictions can help farmers optimize resource allocation and reduce post-harvest losses.

1.3.8 Supply Chain Optimization

The project aims to create direct farmer-to-consumer platforms that minimize post-
harvest losses and enhance market access for farmers. By utilizing data analytics to
understand market demand and pricing trends, the system will empower farmers to make
informed decisions about when and how to sell their produce. This will not only improve
farmers’ profitability but also contribute to food security by reducing waste in the supply
chain.

1.3.9 Accessibility
Recognizing the diverse technological capabilities of farmers, AgriSense AI will prioritize
user-friendly mobile and web interfaces that provide real-time data insights. The platform
will be designed to be accessible to farmers of all backgrounds, ensuring that they can
easily navigate the system and benefit from its features. Training programs and resources
will also be provided to enhance digital literacy among farmers.
Through these focus areas, AgriSense AI aims to create a holistic solution that ad-
dresses the multifaceted challenges faced by the agricultural sector. By empowering farm-
ers with the tools and knowledge they need, the project seeks to foster a more sustainable
and resilient agricultural landscape.

1.4 Motivation
1.4.1 Why Sustainable Farming?
Sustainable agriculture is vital for addressing the growing global food demand while con-
serving essential resources for future generations. As the world population continues
to rise, the pressure on agricultural systems to produce more food increases significantly.
Sustainable farming practices not only aim to meet this demand but also focus on preserv-
ing the environment, enhancing soil health, and reducing the carbon footprint associated
with agricultural activities.

3
 The motivation behind AgriSense AI stems from the urgent need to modernize farm-
ing methods to ensure food security and environmental conservation. By promoting
sustainable practices, farmers can mitigate the adverse effects of climate change, such as
droughts and floods, which threaten crop production. Additionally, sustainable farming
enhances biodiversity, improves ecosystem services, and fosters resilience against pests
and diseases, ultimately leading to more stable food systems.

1.4.2 Role of Technology in Agriculture

Advancements in technology, particularly in AI, remote sensing, and the Internet of
Things (IoT), have the potential to revolutionize agriculture by providing precise, data-
driven solutions. These technologies enable farmers to monitor their fields in real-time,
analyze vast amounts of data, and make informed decisions that enhance productivity
while minimizing environmental impact.
AI algorithms can process data from various sources, including weather forecasts,
soil sensors, and satellite imagery, to generate actionable insights. For example, AI
can identify patterns in crop growth and predict potential issues before they escalate,
allowing farmers to take proactive measures. Remote sensing technology provides valuable
information about crop health, soil moisture levels, and nutrient status, enabling targeted
interventions that optimize resource use.
IoT devices, such as soil moisture sensors and weather stations, facilitate continuous
monitoring of agricultural conditions. This real-time data collection allows farmers to
adjust their practices dynamically, ensuring that they apply the right amount of water,
fertilizers, and pesticides at the right time. By leveraging these technologies, AgriSense
AI aims to create a more efficient and sustainable agricultural ecosystem.

1.5 Implementation Strategy

1.5.1 Data Collection and Analysis
The success of AgriSense AI relies heavily on the collection and analysis of diverse
datasets. The project will utilize satellite imagery, IoT sensor data, and historical agri-
cultural records to create a comprehensive database that informs decision-making.
Data collection will involve partnerships with satellite imaging companies and agricul-
tural research institutions to access high-resolution imagery and relevant datasets. IoT
devices will be deployed in fields to gather real-time data on soil moisture, temperature,
and other critical parameters. This data will be processed using advanced analytics and
machine learning algorithms to derive insights that are tailored to the specific needs of
farmers.

1.5.2 User Engagement and Training

To ensure the successful adoption of AgriSense AI, user engagement and training will be
integral components of the implementation strategy. The project will conduct workshops
and training sessions to educate farmers about the platform’s features and benefits.
User feedback will be actively sought to refine the platform and ensure it meets the
needs of its users. By fostering a collaborative environment, AgriSense AI aims to build
trust and encourage farmers to embrace technology as a means to enhance their practices.

4
1.5.3 Partnerships and Collaborations
Collaboration with various stakeholders, including government agencies, agricultural or-
ganizations, and technology providers, will be essential for the project’s success. Part-
nerships will facilitate knowledge sharing, resource mobilization, and the development of
best practices in sustainable agriculture.
AgriSense AI will also seek to engage with local communities to understand their
unique challenges and tailor solutions accordingly. By working together, stakeholders
can create a supportive ecosystem that promotes sustainable farming practices and drives
positive change in the agricultural sector.

1.6 Conclusion
In conclusion, AgriSense AI represents a transformative approach to modern agriculture,
addressing the pressing challenges of sustainability, resource management, and food se-
curity. By harnessing the power of AI, remote sensing, and IoT, the project aims to
empower farmers with the tools and knowledge they need to thrive in an increasingly
complex agricultural landscape.
The integration of technology into farming practices not only enhances productivity
but also promotes environmental stewardship and resilience. As the agricultural sec-
tor continues to evolve, initiatives like AgriSense AI will play a crucial role in shaping
a sustainable future for agriculture, ensuring that it can meet the needs of a growing
population while preserving the planet for generations to come.

5
Chapter 2

Objectives

2.1 Primary Objectives

The primary objectives of this project focus on addressing critical challenges in agricul-
ture through innovative AI-driven solutions. These objectives aim to empower farmers,
optimize resource utilization, and enhance productivity while maintaining sustainability.
Below are the detailed primary objectives:
• Develop an AI-Based Advisory System for Crop Management and Soil
Health Monitoring: The core goal is to create a comprehensive system that
provides actionable insights on crop health, soil conditions, pest infestations, and
disease outbreaks. By leveraging satellite data, IoT sensor readings, and machine
learning algorithms, the system will guide farmers in making informed decisions,
ultimately improving yield and reducing risks.
• Optimize Irrigation and Fertilizer Usage to Reduce Costs and Environ-
mental Impact: Efficient resource management is a critical aspect of sustainable
farming. The system will use real-time data and AI models to generate precise irri-
gation and fertilization schedules. This will help minimize water wastage, prevent
overuse of fertilizers, and reduce environmental harm caused by runoff and leaching.
• Enhance Decision-Making Using Real-Time Visualizations and Dash-
boards: The project aims to deliver insights through intuitive dashboards and
visual tools. Farmers and agronomists will have access to spatial heatmaps, trend
analyses, and real-time alerts, enabling quicker and more effective decision-making.
This approach ensures that complex data is simplified and presented in a user-
friendly manner.

2.2 Secondary Objectives

The secondary objectives are supplementary goals that align with the broader vision of
promoting sustainable and efficient farming practices. These objectives aim to expand the
system’s capabilities, improve user adoption, and enhance the agricultural value chain.
Below are the detailed secondary objectives:
• Integrate Remote Sensing Data for Advanced Crop Health Assessments:
Leveraging data from remote sensing platforms, including multispectral and hyper-
spectral satellite imagery, the system will perform detailed crop health analyses.

6
 Parameters such as NDVI, EVI, and canopy temperature will be used to detect
stress factors like water deficiency, pest infestations, and nutrient imbalances at an
early stage.

• Train Farmers in Using AI-Driven Tools for Sustainable Farming: Bridg-

ing the technology gap is essential for widespread adoption of AI-based tools. The
project includes organizing workshops, creating multilingual guides, and offering
mobile-friendly tutorials to ensure farmers of varying literacy levels can effectively
use the system. Training modules will also focus on sustainable farming practices
and efficient resource management.

• Promote Direct-to-Consumer Sales for Reducing Post-Harvest Losses:

To support farmers in maximizing their profits, the system will facilitate direct-
to-consumer (D2C) sales by providing market insights and buyer connections. By
reducing reliance on intermediaries, farmers can achieve better price realization, and
post-harvest losses due to delays or storage issues can be minimized. Additionally,
integrating predictive analytics for harvest planning will help farmers align their
supply with market demand.

2.3 Alignment with Sustainable Development Goals

The outlined objectives contribute directly to several United Nations Sustainable Devel-
opment Goals (SDGs), including:

• SDG 2: Zero Hunger - By improving crop yields and reducing losses, the project
supports global food security.

• SDG 6: Clean Water and Sanitation - Optimized irrigation practices will

conserve water resources and protect water quality.

• SDG 13: Climate Action - Efficient resource utilization and sustainable farming
practices will reduce the carbon footprint of agricultural activities.

• SDG 12: Responsible Consumption and Production - The project pro-

motes sustainable consumption of fertilizers and pesticides while reducing food
waste through better planning.

2.4 Conclusion
The primary and secondary objectives of this project are designed to tackle the multi-
faceted challenges faced by modern agriculture. By integrating cutting-edge AI technolo-
gies with traditional farming practices, the system aims to empower farmers, enhance
productivity, and promote sustainability. Through these objectives, the project not only
addresses immediate agricultural challenges but also contributes to the broader goals of
environmental conservation and economic development.

7
Chapter 3

System Design and Architecture

3.1 System Overview

AgriSense AI’s system architecture integrates satellite and UAV (drone)-based inputs
with real-time sensor data and AI models to generate field-specific recommendations.
The architecture is composed of the following layers:

• Data Ingestion Layer: Gathers satellite TIFF files (Sentinel, Landsat), UAV
multispectral imagery (RGB, NIR, Red Edge bands), and IoT data (soil moisture,
pH, temperature). UAV images in TIFF/JPG format undergo orthomosaic gener-
ation and band extraction.
• AI Processing Layer: Uses CNNs and statistical models to:
– Generate vegetation indices: NDVI, NDRE, OSAVI, GNDVI.
– Perform spatial classification using color thresholds.
– Predict yield and water requirements.
• User Interface Layer: Outputs include KMZ files, NDVI heatmaps, advisory
notifications, and dashboards available via mobile and web interfaces.

3.2 Workflow
AgriSense AI follows a five-step processing flow, enhanced to include UAV imagery pro-
cessing for precision agriculture.

1. Data Collection: Ingests satellite images, UAV-captured TIFFs/JPGs, IoT sensor

data, and weather APIs.
2. Preprocessing: Normalization, cloud masking, orthomosaic stitching, and band
splitting.
3. Vegetation Index Calculation (New): Using UAV multispectral data, com-
putes NDVI, NDRE, GNDVI, and OSAVI. Health zones are color-mapped and
classified using thresholds.
4. AI Processing: Applies LSTM for soil moisture forecasting, CNNs for leaf health
detection, and XGBoost for yield estimation.

8
 5. Insight Generation and Feedback: Exports KMZ maps, sends SMS alerts, and
updates dashboards. Farmer feedback loops refine future advisories.

UAV Imagery Preprocessing Vegetation Indices

(NIR, Red, Red Edge) (Band Splitting, Normalization) (NDVI, NDRE, GNDVI)
Dashboards + KMZ
Mobile Alerts + Reports

Satellite Imagery Cloud Removal

(Sentinel, Landsat) and GeoCorrection

AI Models
(CNN, XGBoost, LSTM)

IoT Sensors
(Moisture, pH, EC)

Figure 3.1: System Architecture for AgriSense AI Platform

3.3 Data Pipeline

UAV / Drone Preprocessing

Images (Band Split-
(TIFF, JPG, NIR) ting, NDVI)
AI/ML Models Dashboards, KMZ
(CNN, LSTM, Health Re-
XGBoost) ports, Alerts

Satellite Images
(Sentinel, Landsat)

IoT Sensors Sensor Fusion

(Moisture, + Feature En-
pH, EC) gineering

Figure 3.2: Data Flow Pipeline in AgriSense AI

3.4 System Features and Scalability

• Real-Time Analysis: The system is designed for real-time data processing, en-
suring that users receive timely alerts and recommendations.
• Scalable Architecture: Built using cloud platforms like AWS and Google Cloud,
the system can handle large-scale data from multiple sources and users simultane-
ously.
• Modular Design: Each layer operates independently, enabling easy updates and
the addition of new features.
• Data Security: All user data is encrypted and stored securely, with access control
mechanisms to protect sensitive information.

9
3.5 Conclusion
The AgriSense AI system’s design and architecture exemplify a robust, scalable, and user-
centric approach to modern agriculture. By integrating advanced AI/ML techniques with
geospatial and IoT technologies, the platform delivers precise, actionable insights that em-
power farmers to enhance productivity and sustainability. The seamless workflow ensures
that data flows smoothly from acquisition to actionable insights, making AgriSense AI a
comprehensive solution for addressing the challenges of contemporary farming.

10
Chapter 4

Project Modules and Features

4.1 Data Acquisition and Integration

The system integrates three core data sources:

• Satellite Imagery: Uses Sentinel-2 and Landsat for NDVI and EVI computation
over time.

• UAV Drone Imagery: High-resolution TIFF and JPGs with NIR, Red, and Red
Edge bands are used to calculate vegetation indices. Drone-collected data supports
field-level classification.

• Sensor & Weather Data: IoT devices and real-time APIs provide in-situ soil
readings and weather forecasts.

TIFF images are processed through band-splitting, normalized reflectance compu-

tation, and vegetation index mapping. These are later merged into orthomosaics, and
color-classified overlays are generated using AI.

4.2 Crop and Soil Health Monitoring

Crop stress and growth are tracked using:

• NDVI: Measures greenness. Higher NDVI indicates healthier biomass.

• GNDVI: Highlights nitrogen deficiencies.

• NDRE: Suitable for late-stage crops.

• OSAVI: Adjusts for bare soil in early-stage crops.

UAV images are tiled and each tile is assigned a health class (Healthy, Moderate,
Stressed) based on threshold rules. Color heatmaps are exported as images and KMZ
overlays.
Soil pH, nutrient, and moisture readings are also collected via sensors, enabling hybrid
AI diagnostics from both aerial and ground views.

11
 NDVI ¡ 0 (Water/Soil)

NDVI 0.0 – 0.3 (Stressed)

NDVI 0.3 – 0.5 (Moderate)

NDVI 0.5 – 0.7 (Healthy)

NDVI ¿ 0.7 (Very Healthy)

Figure 4.1: NDVI Value Ranges and Corresponding Health Classifications

NDVI Tiled Heatmaps

Assign GPS Coordinates

KMZ File Generation

Google Earth Overlay

Figure 4.2: KMZ Generation Workflow in AgriSense AI

Healthy
Moderate
Stressed
Bare Soil
Water

Figure 4.3: Simulated KMZ NDVI Overlay for UAV-captured Field

12
4.3 Irrigation Scheduling
An AI-based irrigation advisory system will help farmers optimize water usage. Time-
series forecasting models such as Long Short-Term Memory (LSTM) networks will analyze
real-time soil moisture levels, weather data, and crop water requirements to generate
precise irrigation schedules. This will ensure efficient water usage, reduce wastage, and
enhance crop health.
The system will also leverage weather APIs and integrate rainfall predictions to ad-
just irrigation schedules dynamically, accounting for real-time weather fluctuations. By
analyzing historical irrigation data, the system will provide insights into long-term water
management strategies for sustainable agriculture.

4.4 Yield Prediction

Machine learning models will be trained using historical yield data, satellite imagery, and
real-time weather data to predict crop yields accurately. Gradient Boosting algorithms
like XGBoost will analyze key factors such as soil health, weather patterns, and crop
variety to provide actionable insights.
Yield prediction will empower farmers to plan planting and harvesting schedules ef-
fectively, allocate resources efficiently, and make informed decisions regarding market
supply. By incorporating data from NRSC and historical yield trends, the platform will
continuously improve its predictive accuracy.

4.5 Crop and Irrigation Advisory

The platform will recommend optimal crops based on soil type, nutrient availability, and
weather conditions. Decision Tree and Random Forest models will analyze historical crop
cycles and nutrient uptake patterns to provide tailored recommendations. Farmers will
receive crop-specific advisory information, including best planting practices and nutrient
requirements.
Additionally, the system will generate irrigation schedules using soil moisture data,
crop water requirements, and weather forecasts. These schedules will help farmers main-
tain optimal hydration levels for their crops, reducing water stress and improving yields.

4.6 Supply Chain Optimization

AgriSense AI will address post-harvest challenges by optimizing the agricultural supply
chain. The platform will include the following features:

• Direct Farmer Sales: A marketplace feature will enable farmers to list and sell
their produce directly to consumers. Integration with e-commerce platforms and
payment gateways will ensure seamless transactions.

• Logistics Recommendations: The system will suggest optimal transportation

routes and storage methods to minimize post-harvest losses. By analyzing historical
logistics data, it will provide actionable recommendations to enhance profitability.

13
4.7 Tech Stack Recommendations
The technology stack for AgriSense AI is designed to support high-resolution image pro-
cessing, machine learning, real-time geospatial rendering, and farmer-friendly interfaces.

• Frontend: ReactJS for dynamic web UIs, Bootstrap for responsiveness, and LeafletJS
for rendering geospatial maps.

• Backend: Python (Flask-based microservices) for ML processing, and Node.js for

asynchronous APIs.

• Machine Learning Libraries: TensorFlow and PyTorch for deep learning tasks
(e.g., CNNs for disease classification), Scikit-learn and XGBoost for traditional ML,
OpenCV for image segmentation.

• Geospatial Processing: Rasterio for TIFF handling, GDAL and OpenCV for
band-level manipulations, QGIS for spatial analysis, and simplekml for KMZ gen-
eration.

• Database Systems: PostgreSQL with PostGIS extension for spatial storage, and
MongoDB for metadata and logs.

• Cloud Platforms: Google Cloud (GCS, BigQuery) and AWS (S3, Lambda) for
scalable processing and storage.

• Image Libraries: OpenCV, Rasterio, PIL

• Analysis: NumPy, SciPy, Matplotlib

• Mapping: simplekml, GeoPandas

• Data Storage: TIFF, PNG, CSV, KMZ

• Visuals: Real-time NDVI maps, health overlays, and KMZ file exports

4.8 AI Algorithms and Models

AgriSense AI integrates multiple AI/ML models depending on data types and use cases:

• Convolutional Neural Networks (CNNs): Used for classifying high-resolution

UAV imagery tiles into vegetation health categories using texture, shape, and color
patterns.

• XGBoost: Applied for regression-based yield prediction using vegetation indices,

soil characteristics, and historical crop data.

• LSTM Networks: Used to forecast soil moisture and weather parameters for
optimizing irrigation cycles.

• Random Forest Classifiers: Used for crop type recommendation and disease
susceptibility prediction using past performance data.

14
 • NDVI-Based Threshold Classifier: A rules-based system that uses calculated
NDVI values to classify crop regions as:

– Healthy (NDVI ¿ 0.6)

– Moderate (NDVI between 0.3 and 0.6)
– Stressed (NDVI ¡ 0.3)

4.9 Additional Features

To further enhance the platform, the following features will be incorporated:

• Community Forum: A space for farmers to share insights, solutions, and expe-
riences.

• Educational Resources: Access to videos, tutorials, and guides on best agricul-

tural practices.

• Mobile App: A simplified version of the platform for real-time monitoring and
advisory on-the-go.

4.10 Development Phases

The project will be executed in several phases to ensure systematic development:

• Requirement Analysis: Finalizing data sources and project scope (2-3 weeks).

• Prototype Development: Developing a basic version with core functionalities (1

month).

• Core Feature Integration: Adding soil and crop health monitoring, yield pre-
diction, and supply chain modules (2-3 months).

• Testing and Deployment: Ensuring reliability and scalability through rigorous

testing (1-2 months).

• Maintenance and Updates: Continuous improvements based on user feedback

and technological advancements (Ongoing).

4.11 NDVI and Vegetation Index Module

This module is central to AgriSense AI’s ability to assess crop health using drone imagery.
By analyzing reflectance in various spectral bands (NIR, Red, Green, Red Edge), we
compute vegetation indices that help classify plant health.

• NDVI (Normalized Difference Vegetation Index): Measures overall green-

ness and plant vigor.
N IR − Red
NDVI =
N IR + Red

15
 • GNDVI (Green NDVI): Focuses on nitrogen content in crops.

N IR − Green
GNDVI =
N IR + Green

• NDRE (Red Edge NDVI): Sensitive to chlorophyll in mature leaves.

• OSAVI (Optimized Soil Adjusted VI): Adjusts for background soil interfer-
ence.

The system uses Python libraries such as OpenCV, Rasterio, and NumPy to read georef-
erenced TIFF images, normalize band data, and apply formulas. The results are then:

• Masked to filter non-vegetation areas,

• Color-mapped to indicate health classes (e.g., green for healthy, red for stress),

• Exported as KMZ overlays for use in Google Earth,

• Aggregated into statistical reports for per-plant diagnostics.

16
Chapter 5

Literature Review

5.1 Existing Applications of AI in Agriculture

Artificial Intelligence (AI) has revolutionized agriculture by introducing precision farming,
crop health monitoring, irrigation management, and pest control. Precision farming
leverages AI to analyze soil data, weather patterns, and crop conditions, enabling farmers
to apply resources such as water, fertilizers, and pesticides more efficiently [1]. Examples
include AI-driven systems like Blue River’s ”See Spray,” which uses computer vision to
identify weeds and target them with precision herbicide application [2].
Despite these advancements, most solutions cater to large-scale commercial farms
and are inaccessible to smallholder farmers due to cost, technological complexity, and
lack of localized customization [3]. Furthermore, many solutions lack robust mechanisms
to process localized data, which limits their accuracy in diverse agricultural conditions
[?]. This project, AgriSense AI, aims to bridge these gaps by developing an intuitive,
affordable platform tailored to meet the needs of smallholder farmers globally.

5.1.1 NDVI-Based Plant Health Distribution

Healthy (50%)
Moderate (20%)

Stressed (20%)
Bare Soil/Water (10%)

Figure 5.1: NDVI-Based Plant Health Classification Distribution

17
UAV-Based Remote Sensing vs Satellite Imaging
While traditional satellite platforms like Sentinel-2 and Landsat provide periodic cover-
age, they are often limited by cloud cover, coarse resolution, and fixed revisit schedules.
In contrast, UAVs (Unmanned Aerial Vehicles) offer:

• High spatial resolution (up to 2–5 cm per pixel),

• Flexible and frequent deployments,

• Real-time analysis capabilities for localized crop stress detection.

AgriSense AI integrates both sources—combining large-area satellite imagery for broad

trends with detailed UAV imagery for on-the-ground decision-making. This hybrid ap-
proach improves both scalability and precision.
Combining remote sensing with AI models has enabled breakthroughs in agricultural
productivity. For instance, convolutional neural networks (CNNs) have been used to
process satellite imagery for crop classification and yield prediction [?]. Similarly, GIS
platforms integrated with AI allow for precise mapping of soil nutrient levels, irrigation
zones, and pest outbreaks, enabling farmers to make informed decisions [6].
However, integrating these technologies into user-friendly platforms accessible to farm-
ers remains a challenge. AgriSense AI addresses this gap by providing localized insights
through satellite data processing and AI-based recommendations.

150
150
130
120
Number of Plants

100
100
70

50 40

0
Healthy Moderate Stressed
Plot A Plot B

Figure 5.2: Comparison of Vegetation Health Classes Between Plot A and Plot B

5.2 Remote Sensing and GIS in Agriculture

Remote sensing provides non-invasive monitoring of plant growth, water stress, and pest
activity. AgriSense AI leverages both satellite and UAV-based sources.
Comparison between UAV and Satellite Platforms:

• Satellite Imagery (e.g., Sentinel-2, Landsat-8):

18
 – Spatial resolution of 10–30 meters.
– Limited by cloud cover and revisit cycles (5–10 days).
– Suitable for regional-scale monitoring.

• Drone/UAV Imagery:

– Spatial resolution of 2–5 cm.

– On-demand flights and near-instant data collection.
– Enables per-plant analysis and high-resolution NDVI, GNDVI, NDRE map-
ping.

AgriSense AI fuses both sources. Satellite imagery offers macro trends while drones
provide field-specific intelligence, enabling a hybrid AI workflow. KMZ exports from UAV
images allow stakeholders to overlay findings in Google Earth with high precision.

5.3 Addressing Knowledge Gaps

Despite significant progress, several knowledge gaps persist in the deployment of AI and
remote sensing for agriculture. Smallholder farmers often lack the technical knowledge
to interpret and utilize complex AI systems. Moreover, high costs and infrastructure
limitations hinder widespread adoption [?].
AgriSense AI aims to address these gaps by:

• Developing cost-effective tools that are scalable and adaptable to various farming
conditions.

• Offering localized AI models trained on region-specific data to improve accuracy

and relevance.

• Providing training and capacity-building programs for farmers to ensure effective

use of the technology.

• Emphasizing sustainability by incorporating environmental considerations into AI

recommendations.

The project also focuses on the long-term sustainability of agriculture by addressing

challenges such as resource scarcity, climate change adaptation, and biodiversity conser-
vation [11].

5.4 Related Work and Recent Advances

Recent advancements in AI and IoT for agriculture have further highlighted their poten-
tial. For instance:

• AI for Pest Control: AI models, such as those using deep learning, can identify
pests in real-time using smartphone cameras or drones. Research by [7] shows a
90% accuracy in pest detection using convolutional neural networks.

19
 • Smart Irrigation Systems: IoT-enabled sensors combined with AI optimize wa-
ter usage by predicting soil moisture levels and rainfall patterns [8].

• Crop Disease Detection: Deep learning-based image analysis has shown remark-
able results in detecting crop diseases from leaf images, with systems like PlantVil-
lage achieving high accuracy rates [9].

• Yield Prediction Models: AI models such as random forests and recurrent neural
networks (RNNs) have been used for yield prediction based on climatic and soil data
[10].

Despite these advancements, scaling such solutions to smallholder farms remains a

critical research challenge. AgriSense AI proposes an accessible and adaptable framework
to address these limitations.

20
Chapter 6

Future Work

6.1 Phase 2 Development

Phase II of AgriSense AI will focus on deeper predictive analytics, automation, and user
engagement features:

• Deep Learning Leaf Disease Detection: Expand use of CNNs on UAV-acquired

visible/NIR images to classify leaf-level fungal, bacterial, and viral infections.

• Temporal Vegetation Analysis: Allow time-series plots of NDVI, GNDVI, and

LCI for each crop zone to monitor stress trends over time.

• Interactive Dashboards: Real-time updates on crop health metrics, irrigation

needs, and yield predictions.

• KMZ Overlay Engine (Advanced): Generate multi-layered Google Earth files

showing vegetation class, sensor readings, and historical yield data.

• Voice Assistant + Mobile App: Enable farmers to receive voice-based sum-

maries in their local language, and to ask questions about crop recommendations
or upcoming weather alerts.

Leaf Disease Detection(CNN + UAV RGB)

Temporal Index Monitoring(NDVI Trend Lines) Dashboards + Alerts

Multi-layer KMZ Overlays Voice Assist in Mobile App

Figure 6.1: Phase II Roadmap for AgriSense AI

21
 1
NDVI Value

0.5
Plot A
Plot B
0
1 2 3 4 5
Flight Date

Figure 6.2: Simulated NDVI Trendlines for Two Field Plots

1
Flight 1 (Early Stage)
Flight 2 (2 Weeks Later)
0.8
Average NDVI

0.6

0.4

0.2
A1 A2 A3 A4 A5
Sub-Plot ID

Figure 6.3: NDVI Evolution Over Time – Flight 1 vs Flight 2

22
6.2 Evaluation and Impact Assessment
A comprehensive evaluation framework will be established to measure the impact of
AgriSense AI on various metrics, including:

• Agricultural Productivity: Monitoring changes in crop yields and resource effi-

ciency.

• User Satisfaction: Conducting surveys and interviews to assess usability and

farmer satisfaction.

• Environmental Sustainability: Measuring reductions in water and fertilizer us-

age, as well as carbon footprint.

• Economic Impact: Evaluating improvements in farmer income and cost savings.

The evaluation process will involve both quantitative metrics (e.g., yield data, resource
usage) and qualitative feedback (e.g., farmer testimonials).

23
Chapter 7

Conclusion

AgriSense AI represents a holistic and transformative approach to modern agriculture by

addressing the key challenges of accessibility, scalability, and sustainability. By leveraging
AI, remote sensing, and IoT, the project empowers farmers with actionable insights and
user-friendly tools.
The initiative not only enhances agricultural productivity but also promotes sustain-
able practices and resilience to environmental challenges. With the planned Phase 2
developments and evaluation framework, AgriSense AI aims to make a lasting impact on
the agricultural sector, ensuring food security and environmental stewardship for future
generations.

24
Bibliography

25
Bibliography

[1] J. Smith, ”Precision Agriculture Using Artificial Intelligence,” Agricultural Technol-

ogy Journal, vol. 34, no. 2, pp. 56-70, 2022.

[2] Blue River Technology, ”See & Spray Technology Overview,” 2023. [Online]. Avail-
able: https://www.bluerivertechnology.com.

[3] M. Patel et al., ”AI Challenges for Smallholder Farmers,” Global Agriculture, vol.
12, pp. 44-50, 2021.

[4] R. Tanaka et al., ”Applications of Remote Sensing in Modern Agriculture,” Remote

Sensing Letters, vol. 15, no. 5, pp. 345-360, 2020.

[5] A. Gomez et al., ”NDVI for Crop Health Monitoring,” International Journal of
Agriculture, vol. 19, pp. 89-95, 2021.

[6] P. Chandra, ”GIS in Precision Farming,” Journal of Agricultural Science, vol. 47,
no. 3, pp. 112-123, 2019.

[7] L. Nguyen, ”Deep Learning for Pest Control in Agriculture,” AI in Agriculture, vol.
8, pp. 34-40, 2022.

[8] K. Sharma et al., ”IoT-Enabled Smart Irrigation Systems,” IoT and Agriculture, vol.
3, pp. 78-85, 2020.

[9] PlantVillage, ”AI for Crop Disease Detection,” 2022. [Online]. Available: https:
//www.plantvillage.psu.edu.

[10] M. Zhou et al., ”Yield Prediction Using RNN Models,” AI for Sustainable Agricul-
ture, vol. 7, no. 1, pp. 13-22, 2021.

[11] B. Lee, ”Sustainable Agriculture Through AI,” Sustainability Journal, vol. 29, pp.
215-229, 2020.

“‘Additional sections and details will be added for Phase 2/2nd review. “‘

26
Biodata
Name: Aditya Khamitkar
Roll Number: 22MIP10006
Email: [email protected]
Contact Number: +91-9014810008
Program: Integrated M.Tech Specialization in
Computational and Data Science
Year: Third Year
University: VIT Bhopal University
Skills and Interests: Image Processing, Artificial
Intelligence, Machine Learning, and Satellite Data
Analysis
Notable Achievements: Intern at ISRO(SDSC SHAR)
focusing on PSLV flight safety, Mission Computers, and
Specialist Display Systems.
Hobbies: Swimming, Bird Watching, Aerospace and
exploring AI-driven technologies.

Name: Devendra Patil

Roll Number: 22MIP10063
Email: [email protected]
Contact Number: 7798766686
Program: Integrated M.Tech Specialization in
Computational and Data Science
Year: Third Year
University: VIT Bhopal
Skills and Interests: Artificial Intelligence, Machine
Learning, SQL, Database Management
Hobbies: Reading Books, sports especially Cricket.

27
 Name: Aadish Pankaj Dabhade
Roll Number: 22MIP1005
Email: [email protected]
Contact Number: +91 73876 59119-
Program: Integrated M.Tech Specialization in
Computational and Data Science
Year: Year 3
University: VIT Bhopal University
Skills and Interests:Data Analysis and Processing,
Java Development
Hobbies: Sports like Skating.

Name: Abhay Kumar

Roll Number: 22BCE11152
Email: [email protected]
Contact Number: +91-8306580122
Program: B.Tech Computer Science
Year: Third Year
University: VIT Bhopal University
Skills and Interests: Artificial Intelligence and
Machine Learning, Database Management and Analysis,
Image Processing
Hobbies: Literature reading, Latest Tech Enthusiast,
Weight Lifting.

28
 Name: Pranav Sanjay Jahagirdar
Roll Number: 22BCY10233
Email: [email protected]
Contact Number: +91-74988-82171
Program: B.Tech in Cybersecurity
Year: Third Year
University: VIT Bhopal University
Skills and Interests: Placeholder Skills 5
Hobbies: Chess, ethical hacking, coding, photography.

Name: Angiras Devendra Bawkar

Roll Number: 22BAS10004
Email: [email protected]
Contact Number: +91-8850276533
Program: B.Tech Aerospace Engineering
Year: Third Year
University: VIT Bhopal University
Skills and Interests: CAD, Structures, IoT, Drones,
Microcontrollers, Flight Mechanics
Hobbies: Reading about Tech.

Name: Vaibhavi Rajguru

Roll Number: 22BAS10006
Email: [email protected]
Contact Number: +91-9421592731
Program: B.Tech in Aerospace Engineering
Year: Third Year
University: VIT Bhopal University
Skills and Interests: Placeholder Skills 5
Hobbies: Swimming, stargazing, playing violin,
painting.

29