0% found this document useful (0 votes)
10 views

Iot Project

This project report presents an IoT-enabled system for diagnosing plant diseases using image processing and machine learning, specifically Convolutional Neural Networks (CNNs). It aims to address the challenges of traditional manual disease detection methods by providing automated, real-time monitoring and analysis to improve agricultural yield. The report outlines the design, implementation, and future work needed to enhance the scalability and accuracy of the system in diverse agricultural environments.

Uploaded by

Opendra Yadav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
10 views

Iot Project

This project report presents an IoT-enabled system for diagnosing plant diseases using image processing and machine learning, specifically Convolutional Neural Networks (CNNs). It aims to address the challenges of traditional manual disease detection methods by providing automated, real-time monitoring and analysis to improve agricultural yield. The report outlines the design, implementation, and future work needed to enhance the scalability and accuracy of the system in diverse agricultural environments.

Uploaded by

Opendra Yadav
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

IoT-Enabled Smart Plant Disease

Diagnosis Using Image Processing and


Machine Learning

A PROJECT REPORT

Submitted by

Muskan (22BCS13402)
Rajvir Singh (22BCS13717)
Abhishek(22bcs13366)

in partial fulfillment for the award of the degree of

BACHELOR OF ENGINEERING

IN

COMPUTER SCIENCE

Chandigarh University

April & 2025

1
2
TABLE OF CONTENTS

CHAPTER 1: - INTRODUCTION ........................................................ 4


1.1. Identification of client ......................................................................... 4
1.2. Identification of problem ......................................................................4
1.3. Identification of Tasks ......................................................................... 5
1.4. Organisation of the Report .................................................................... 5
CHAPTER 2: - Literature Survey ............................................................7
2.1. Timeline of the reported problem ......................................................... 7
2.2. Existing solutions .................................................................................. 8
2.3. Bibliometric analysis ............................................................................ 9
2.4. Review summary................................................................................... 8
2.5. Problem definition ............................................................................... 11
2.6. Goals/Objective ................................................................................... 13

CHAPTER 3: - DESIGN FLOW/PROCESS ......................................... 15


3.1. Evaluation and selection of specifications/features ............................... 15
3.2. Design Constraints ............................................................................... 16
3.3. Analysis of features and finalization subject to constraints ....................16
3.4. Design flow ...........................................................................................17
3.5. Design selection ................................................................................... 17
3.6. Implementation plan/Methodology........................................................ 18

CHAPTER 4: - RESULTS ANALYSIS ................................................... 19


4.1. Implementation of design using modern technical tools ........................ 19
4.2. Results and testing ................................................................................ 20

CHAPTER 5: - CONCLUSION & FUTURE WORK ........................... 21


5.1. Conclusion............................................................................................ 21
5.2. Future work ...........................................................................................21

REFERENCES .......................................................................................... 22

3
CHAPTER 1

INTRODUCTION

1.1. Identification of Client /Need / Relevant Contemporary issue

The primary clients for this system are farmers, agricultural researchers, and policymakers focused on
improving agricultural yield through technological innovation. Specifically, this IoT-enabled system
addresses the needs of farmers who require fast, accurate, and actionable insights to combat plant
diseases. The agricultural community has long relied on manual techniques to identify plant diseases,
a process that is labor-intensive and prone to error. With the shift towards precision farming, the need
for automated systems has become paramount. By implementing this IoT-based solution, farmers can
monitor large areas of farmland with minimal effort. This system also appeals to agricultural
researchers who aim to optimize plant disease management strategies, ensuring sustainability in
farming practices. Policymakers, on the other hand, can use data generated by this system to shape
policies that mitigate the economic impact of crop losses due to disease. Table 1 below summarizes
the key client groups and their expectations.

TABLE – 1

1.2. Identification of Problem

Agriculture, being one of the most critical sectors globally, faces the challenge of detecting plant
diseases at early stages to avoid crop losses. Traditional methods of disease detection often rely on
human observation, which is not only time-consuming but also limited in terms of scalability. Small-
scale farms may be able to rely on visual checks, but for large-scale farming operations, manual
methods become impractical. Furthermore, human error can lead to misdiagnosis or late identification,
worsening the spread of disease. Climate variability and the presence of multiple disease-causing
pathogens compound the difficulty of early and accurate detection. In this project, we aim to tackle
this issue by integrating IoT devices for real-time image capture and machine learning algorithms,
especially CNNs, for precise disease detection. The automated system ensures scalability, minimizes
human error, and provides immediate feedback, which is crucial in regions where agricultural
productivity is directly tied to economic stability.

4
1.3. Identification of Tasks

Agriculture, being one of the most critical sectors globally, faces the challenge of detecting plant
diseases at early stages to avoid crop losses. Traditional methods of disease detection often rely on
human observation, which is not only time-consuming but also limited in terms of scalability. Small-
scale farms may be able to rely on visual checks, but for large-scale farming operations, manual
methods become impractical. Furthermore, human error can lead to misdiagnosis or late identification,
worsening the spread of disease. Climate variability and the presence of multiple disease-causing
pathogens compound the difficulty of early and accurate detection. In this project, we aim to tackle this
issue by integrating IoT devices for real-time image capture and machine learning algorithms,
especially CNNs, for precise disease detection. The automated system ensures scalability, minimizes
human error, and provides immediate feedback, which is crucial in regions where agricultural
productivity is directly tied to economic stability

1.4. Organization of the Report

• Chapter 2 Literature review, we delve into the historical context and evolution of the problem,
reviewing previous research and existing solutions related to the identified problem. We
conduct a bibliometric analysis to analyze relevant literature and summarize the findings of
the literature review. Based on these insights, we refine the problem statement and define the
goals and objectives of the project.

• Chapter 3 Design Flow or Process, we evaluate and select the specifications or features
required for the solution, taking into account any constraints that may impact the design
process. We analyze the identified features and finalize the design considering these
constraints, outlining the flow of the design process from conceptualization to
implementation. Finally, we select the final design and outline the plan for its implementation.

• Chapter 4 Results Analysis and Validation, we describe how the solution was implemented
and present the results of the analysis. Finally, in the Conclusion and Future Work (Chapter
5), we summarize the key findings of the project, discuss implications, and suggest avenues
for future research and improvement.

• Chapter 5 Conclusion and Future work, Summarizes the key findings of the project,
discusses implications, and suggests avenues for future research and improvement.

5
CHAPTER 2

LITERATURE REVIEW/BACKGROUND STUDY

2.1. Timeline of the reported problem

The problem of plant disease identification has been an ongoing challenge in agriculture, dating back
to early 20th-century practices. Initially, manual observation by agricultural experts was the only
available method for detecting diseases, which often resulted in delayed identification and higher crop
losses. In the 21st century, advancements in technology, particularly in image processing and machine
learning, began to offer promising solutions for automated plant disease diagnosis. From 2010
onwards, research on deep learning models like CNNs for plant disease classification gained traction,
with several studies proving that machine learning could significantly outperform traditional methods.
The integration of IoT into agriculture around 2015 marked a turning point, enabling real-time
monitoring and data collection. Today, automated disease detection systems that combine IoT and
deep learning offer not only timely diagnosis but also increased scalability, a crucial factor in large-
scale farming. Figure 2 shows the evolution of plant disease detection technologies, from manual
techniques to IoT-integrated solutions.

Figure :2

7
2.2. Existing solutions

Several existing solutions have been proposed and implemented to address plant disease detection.
Traditional methods rely on physical inspection and manual diagnosis, often conducted by
agronomists or agricultural workers. While somewhat effective, these methods are labor-intensive and
prone to human error, especially for large farming operations. In recent years, automated solutions
have been developed that utilize computer vision, remote sensing, and image processing techniques.
Machine learning, particularly CNNs, has proven effective in classifying plant diseases by analyzing
leaf images. The implementation of IoT further enhances these systems by allowing real-time data
collection, reducing the delay between disease onset and detection. For instance, Asif Iqbal et al.
(2020) reported a system integrating CNNs with IoT devices for real-time detection of diseases like
blight and rust, achieving significant improvements in accuracy. However, the challenge remains in
the scalability and reliability of these systems across different environments and crop types. Table 2
summarizes key solutions in the field and their respective performance metrics.

Table : 2

8
2.3. Bibliometric analysis

A bibliometric analysis of recent research in plant disease detection reveals a significant increase in
publications, particularly those focusing on deep learning and IoT integration. The analysis shows that
over the past decade, the number of studies addressing plant disease management using advanced
technologies has grown exponentially. Keywords such as “CNN,” “IoT,” and “plant disease
diagnosis” dominate the research landscape, highlighting the shift towards automated solutions.
Journals focusing on agricultural engineering, artificial intelligence, and IoT applications are the
primary publishers of these studies. For instance, in 2023 alone, more than 100 papers were published
on the use of IoT-enabled systems for disease detection in crops. Figure 3 below illustrates the number
of publications from 2010 to 2023 related to IoT and deep learning in agriculture, demonstrating the
increasing focus on technological interventions in this sector.

2.4. Review Summary

The literature review highlights several advancements in plant disease detection, with a particular
focus on the role of deep learning models such as Convolutional Neural Networks (CNNs). CNNs
have emerged as the most effective tools for image classification tasks due to their ability to
automatically learn and recognize complex patterns from large datasets of images. In the context of
plant disease detection, CNNs have shown remarkable accuracy in identifying disease symptoms
from leaf images, outperforming traditional machine learning models like Support Vector Machines
(SVMs) and Decision Trees. CNNs can learn intricate features from diseased leaf textures, spots,
and discoloration, which are often too complex for simpler models to accurately capture.

In recent years, studies have demonstrated that CNNs can achieve high classification accuracy,
often above 90%, making them a reliable choice for automated plant disease diagnosis. For
example, CNN models have been trained on datasets like PlantVillage, which contains thousands of
annotated leaf images, to classify diseases such as early blight, late blight, and rust across multiple
crops. These models can discern between subtle differences in disease symptoms, which is crucial
for early intervention. Early detection allows farmers to mitigate damage and reduce the spread of
disease, leading to increased crop yields and improved food security.

Another key advancement discussed in the literature is the integration of Internet of Things (IoT)
devices into plant disease detection systems. IoT devices, equipped with sensors and cameras,
enable continuous, real-time monitoring of crops in the field. This is a significant leap from
traditional methods, which rely on manual inspections that are time-consuming, prone to human
error, and unsuitable for large-scale farming. With IoT integration, farmers can deploy devices
throughout their fields to capture real-time images and environmental data, such as temperature and
humidity, which can influence disease development. These images and data are transmitted to a
central server or cloud-based platform, where machine learning models, particularly CNNs, process
the data and provide instant feedback on the plant’s health status. This allows for timely disease
detection and immediate intervention, such as pesticide application or other treatments, greatly
enhancing the efficiency of disease management practices.

9
Despite these advancements, scalability remains a significant challenge, particularly in diverse and
dynamic environmental conditions. While CNNs have demonstrated high accuracy in controlled
settings or on specific datasets, their performance can degrade when applied in real-world scenarios
with varying lighting conditions, diverse backgrounds, or different crop types. For instance, factors
like shadows, poor lighting, or occlusions in images can reduce the effectiveness of CNNs, leading
to misclassification or false negatives. Moreover, CNNs trained on specific datasets may struggle to
generalize to new crops or diseases that were not part of the training data, limiting the model’s
scalability across different agricultural contexts.

Scalability is further hindered by environmental factors such as soil conditions, moisture levels, and
temperature variations, all of which can affect the visual appearance of disease symptoms on
leaves. These variables introduce noise into the data, making it harder for models to maintain high
accuracy. To address this, existing solutions need refinement to better handle variability in crop
types, disease severity, and environmental conditions. One approach to overcoming these
challenges is through data augmentation techniques, which artificially expand the training dataset
by introducing variations such as rotations, lighting adjustments, and zooms. This helps CNNs
become more robust to environmental variability. Additionally, preprocessing techniques like
normalization and noise reduction can improve the quality of input images, thereby enhancing
model performance.
To further enhance scalability and accuracy, hybrid approaches that combine CNNs with other
techniques, such as clustering, are gaining attention. Clustering techniques can group similar
diseases or crop types based on shared characteristics, enabling the system to classify diseases more
effectively even when encountering unfamiliar conditions. For instance, by clustering diseases with
similar visual symptoms, the system can reduce the complexity of classification tasks and improve
the robustness of the model. This approach allows the system to perform well in diverse agricultural
settings, as it can adapt to new crops and diseases more efficiently.
In conclusion, while deep learning models like CNNs represent a significant advancement in plant
disease detection, challenges related to scalability and environmental variability remain. Integrating
IoT devices for real-time monitoring is a crucial step forward, but more work is needed to refine
these systems for use in dynamic, real-world agricultural environments. Hybrid approaches, such as
combining CNNs with clustering techniques, hold promise for improving both the accuracy and
scalability of plant disease detection systems, ultimately providing farmers with more reliable tools
for disease management and crop protection.

10
2.5. Problem Definition

Despite significant progress in the development of automated plant disease detection systems, several
critical challenges remain unaddressed. Current solutions, while offering valuable advances in
accuracy and efficiency, often face limitations when it comes to scalability and adaptability in real-
world agricultural settings. One of the key gaps in existing systems is their inability to maintain high
accuracy across diverse environmental conditions and varying crop types. Most machine learning
models, particularly those based on deep learning techniques like Convolutional Neural Networks
(CNNs), are trained on specific datasets that are often collected under controlled conditions. This
makes them highly effective in recognizing diseases within the scope of their training data, but less
reliable when applied to broader or more diverse agricultural contexts.

A major limitation of current systems is scalability. As farms grow in size and complexity, disease
detection systems must be able to scale efficiently to monitor large areas continuously without
compromising accuracy. However, many existing solutions are designed for small-scale
implementations and lack the infrastructure or computational efficiency to handle the vast amounts of
data generated by large-scale farming operations. This limits their practical applicability, especially in
regions where agricultural plots are extensive and require comprehensive, real-time monitoring. For
instance, a system that works effectively on a small plot of land may not perform well when deployed
across several hundred acres, leading to delays in disease detection and, consequently, a higher risk of
crop losses.

Another significant issue is the difficulty in maintaining high accuracy in diverse environmental
conditions. Factors such as varying lighting conditions, fluctuating weather patterns, differences in
soil composition, and other environmental variables can affect the appearance of diseased plants,
complicating the machine learning model’s ability to accurately classify diseases. For example, a
disease that manifests with certain visual symptoms under optimal lighting conditions may appear
differently under overcast skies or in shadowed areas of the field. Similarly, differences in background
noise, such as weeds or other crops growing nearby, can introduce confusion for image classification
models. Current systems often struggle to adjust to these variables, leading to inaccuracies in disease
diagnosis. Moreover, many models are trained on data from specific regions or crops, limiting their
ability to generalize across different agricultural ecosystems.

In addition to these issues, while real-time disease monitoring is technically achievable through the
integration of IoT devices, further optimization is required to ensure quick and reliable feedback for
farmers. IoT devices can capture and transmit data in real-time, but the time it takes to process and
analyze this data can vary depending on network speeds, computational power, and the complexity of
the machine learning algorithms in use. For instance, in rural or remote farming areas, where internet
connectivity may be limited or unstable, the system’s ability to provide timely disease diagnosis can
be hindered. This delay in receiving feedback reduces the system’s effectiveness in preventing the
spread of disease and mitigating crop losses. Therefore, optimizing both the data collection and
processing components of the system is crucial to improving its overall performance.

11
The proposed system in this research aims to address these gaps by integrating IoT technology for
real-time data collection with a CNN-based machine learning model that is capable of handling a
variety of crops and environmental conditions. By leveraging IoT devices such as cameras and sensors
placed in the field, the system continuously collects high-resolution images and environmental data,
which are then transmitted to a central server for processing. This real-time data collection allows for
continuous monitoring of crops, ensuring that diseases are detected early and corrective measures can
be implemented immediately.

To enhance the system’s scalability, the model is designed to operate efficiently on cloud-based
platforms, which can handle the large volumes of data generated by IoT devices spread across vast
agricultural areas. The cloud infrastructure ensures that the system can scale up its processing power
as needed, without overburdening the local IoT devices or requiring excessive computational
resources on-site. This allows the system to monitor large-scale farms in real-time, providing timely
feedback to farmers and minimizing the risk of widespread crop loss.

The use of a CNN-based model is key to improving the accuracy of disease detection. CNNs are well-
suited to image classification tasks because they can automatically learn hierarchical patterns from
raw image data, making them ideal for detecting the subtle differences in leaf discoloration, texture, or
other visual symptoms associated with various plant diseases. However, to address the issue of
environmental variability, the system incorporates advanced preprocessing techniques such as data
augmentation and normalization. These techniques help the model become more robust to changes in
lighting conditions, backgrounds, and other environmental factors. Additionally, the system is
designed to be adaptable to multiple crop types by training the CNN on a diverse dataset that includes
images of various crops and diseases from different regions and climates.

Furthermore, the system employs a hybrid approach that combines CNN-based feature extraction with
clustering techniques. This hybrid model allows for the grouping of similar diseases or conditions,
enabling the system to classify new or previously unseen diseases more accurately. By clustering
diseases with similar visual symptoms, the system can improve its generalization capabilities, making
it more effective in handling a wide range of crop types and environmental conditions.

In conclusion, the proposed system offers a robust and scalable solution to plant disease detection by
integrating IoT technology for real-time data collection with a CNN-based machine learning model.
This hybrid approach addresses the limitations of current systems by enhancing scalability, improving
accuracy in diverse environmental conditions, and ensuring timely interventions that minimize crop
losses. Through continuous monitoring and immediate feedback, the system provides farmers with the
tools they need to effectively manage plant diseases and optimize agricultural productivity.

12
2.6. Goals/Objectives

The primary objective of this project is to develop an intelligent, IoT-enabled system that leverages
advanced machine learning models, specifically Convolutional Neural Networks (CNNs), to diagnose
plant diseases in real-time. This system is designed to address the growing need for efficient, accurate,
and scalable solutions in modern agriculture, where timely detection and management of plant
diseases are critical to minimizing crop losses and improving overall productivity. By employing
CNNs, which have proven to be highly effective in image classification tasks, the system aims to
analyze leaf images captured by IoT devices and accurately identify disease symptoms, allowing
farmers to take immediate corrective actions.

A core focus of the project is on real-time disease diagnosis. This means that IoT devices deployed in
agricultural fields will continuously monitor the crops, capturing images and transmitting them to a
central server where the machine learning model processes the data instantly. The system's real-time
capability is crucial for detecting diseases at an early stage, preventing their spread, and reducing the
need for extensive use of pesticides, which can have both economic and environmental drawbacks. By
providing immediate feedback, farmers can make timely decisions regarding disease management,
ultimately improving crop health and yields.

In addition to real-time detection, the system’s secondary objectives are focused on enhancing the
accuracy of disease classification. While traditional methods of plant disease detection, such as
manual inspections, are often prone to human error and require significant expertise, this IoT-enabled
system aims to surpass such limitations. The use of CNNs allows the system to learn and recognize
complex patterns in plant leaves, differentiating between healthy plants and various disease conditions
with high precision. This enhanced accuracy is critical for ensuring that farmers receive reliable
information and can trust the system's recommendations for disease management.

Another significant objective is expanding the scalability of the system to handle large-scale
agricultural operations. Modern farms, especially those spanning vast areas, present unique challenges
for disease monitoring, as manually inspecting crops is time-consuming and labor-intensive. The IoT-
enabled system addresses this by providing a scalable solution that can monitor large fields
continuously without requiring human intervention. IoT devices can be distributed across multiple
locations, allowing the system to gather and process vast amounts of data. The use of cloud-based
infrastructure ensures that the system can scale up its data processing capabilities as needed, handling
data from thousands of IoT devices deployed across different regions. This scalability is crucial for the
system's adoption in large agricultural settings, where timely disease detection and management are
essential for maintaining productivity.

Reducing human intervention in the disease detection process is another key objective. Traditional
methods rely on experts to visually inspect plants and diagnose diseases, which is not only labor-
intensive but also dependent on the availability of skilled professionals. The IoT-enabled system
automates this process, reducing the need for continuous human oversight. Once installed, the system
operates autonomously, capturing and processing data, classifying diseases, and providing actionable
insights directly to the farmer via mobile or web applications. This automation not only reduces the
time and labor required for disease management but also minimizes the risk of human error, leading to
13
more consistent and accurate results.

Moreover, the system is designed to handle a variety of environmental conditions and crop types,
ensuring its broad applicability across different agricultural sectors. Farms located in diverse
geographical regions face different challenges, such as variations in climate, soil conditions, and crop
types. The system's machine learning models are trained on a wide range of datasets that include crops
grown in different environments, making the system robust enough to function effectively in varying
conditions. This adaptability is key to ensuring that the system can be used by farmers worldwide,
regardless of the specific crops they are cultivating or the environmental challenges they face. The
system's ability to operate under different lighting conditions, backgrounds, and environmental
factors, such as temperature and humidity, further enhances its versatility.

In summary, the primary objective of this project is to develop a robust, smart, IoT-enabled system
that leverages CNN-based machine learning for real-time plant disease diagnosis. This system will
enable accurate, timely feedback for farmers, helping them manage diseases more efficiently and
improve crop productivity. By focusing on enhancing classification accuracy, scalability, and reducing
human intervention, the system aims to provide a comprehensive solution that can be deployed across
various agricultural settings, making it a valuable tool in modern precision farming.

14
CHAPTER 3

DESIGN FLOW/PROCESS

3.1. Evaluation & Selection of Specifications/Features

The first step in designing the IoT-enabled plant disease diagnosis system involves evaluating various
system specifications and selecting the most appropriate features for optimal performance. The
primary feature of the system is real-time disease detection through IoT devices equipped with
cameras and sensors. The IoT devices capture high-resolution images of plant leaves, which are then
transmitted to a central server for processing. To ensure high accuracy in disease classification, CNNs
were selected as the primary machine learning model, as they excel in image classification tasks.
Additionally, the system is designed to handle a variety of environmental conditions, including
different lighting scenarios and background noise in the images. To maintain the system's scalability,
it was important to include features that enable it to work with large datasets and multiple crops.
Table 3 outlines the key features evaluated and their selection criteria.

Table 3 : Evaluation of System Specifications and Features

15
3.2. Design Constraints

Several design constraints were considered during the development of the system to ensure it met
performance benchmarks while maintaining feasibility in real-world agricultural settings. One of
the main constraints was the need for a lightweight system that could operate in areas with limited
network connectivity. IoT devices must be capable of capturing images and transmitting them to a
central server in real-time, even in rural areas with poor internet access. Power consumption was
another major constraint, as IoT devices are often deployed in fields for long durations without
access to a stable power source. To address this, the devices were equipped with energy-efficient
processors and solar-powered batteries. Another design constraint was the computational power
required to process large amounts of data in real-time. This was mitigated by utilizing cloud-based
servers, which offload the computational burden from the IoT devices.

3.3. Analysis of Features and finalization subject to constraints

After evaluating the potential features, an analysis was conducted to determine which features could be
implemented within the constraints mentioned. The real-time image processing capability was
finalized as a core feature, as it allows the system to provide timely feedback to farmers. However, to
accommodate the constraint of limited computational resources, feature extraction and image
classification tasks were offloaded to cloud servers. This reduced the processing load on the IoT
devices, allowing them to operate with limited power and network bandwidth. Additionally, data
compression techniques were implemented to reduce the size of the images before transmission,
minimizing the system’s data usage. In terms of environmental adaptability, the system was designed
with flexible image preprocessing techniques, such as normalization and noise reduction, to account
for variations in lighting and background conditions. Table 4 provides a summary of the features
finalized after analyzing the constraints.

Table 4 Features Finalised After Constraints Analysis

16
3.4. Design Flow

The design flow of the system follows a structured sequence of steps, beginning with data collection
and ending with real-time disease diagnosis. The first step involves IoT devices capturing images of
plant leaves, which are then transmitted to a server. On the server, image preprocessing techniques,
such as resizing and normalization, are applied to enhance the image quality. Feature extraction is
performed using CNNs, which identify disease symptoms based on the texture, color, and shape of the
leaf lesions. The system then classifies the disease using clustering techniques combined with machine
learning models. Finally, the results are displayed on a user-friendly interface, providing farmers with
real-time feedback.

3.5. Design selection

Several design options were considered, including different machine learning models and IoT
device configurations. The selection of CNNs as the primary model was based on their proven
effectiveness in image classification tasks. Other models, such as Support Vector Machines
(SVMs), were tested but did not achieve the same level of accuracy. Additionally, the choice to
offload image processing tasks to cloud servers was driven by the need to balance computational
efficiency and power consumption. The IoT devices were selected based on their compatibility with
solar power systems, ensuring long-term sustainability in rural areas. Table 5 compares the
performance of the different models considered for the system.

Table 5 Comparison of Machine Learning For Disease Detection

17
3.6. Implementation plan/methodology

18
CHAPTER 4

RESULTS ANALYSIS AND VALIDATION

4.1. Implementation of design using Modern Technical tools

The system was implemented using modern technical tools, including IoT devices for image capture
and machine learning frameworks such as TensorFlow for disease classification. The use of CNNs
for feature extraction proved highly effective, achieving an accuracy of 90.12% in plant disease
detection. IoT devices deployed in the field continuously monitored crops, capturing real-time
images and transmitting them to the central server for analysis. The system was also integrated with
cloud-based services for data storage and processing, which allowed it to handle large datasets
efficiently. The implementation process involved several technical tools, including Python for image
preprocessing, Keras for model training, and OpenCV for image augmentation. Table 6 outlines the
key technical tools used in the system's implementation.

TABLE 6 Technical tools used in system implementation

4.2. Results and testing

During testing, the system was evaluated on a dataset consisting of thousands of images of plant
leaves affected by various diseases. The system was able to correctly classify plant diseases with an
accuracy of 90.12%, outperforming other models such as Random Forest and SVM. The system’s
precision, recall, and F1 score were also evaluated to measure its effectiveness in disease detection.
Precision refers to the percentage of correct positive predictions, while recall measures the system’s
ability to identify all relevant cases. The F1 score, which balances precision and recall, was
particularly high, indicating that the system performs well in both identifying diseases and
minimizing false positives. Table 7 presents the performance metrics of the system.

19
TABLE 7: Metrics Of System

20
CHAPTER 5 CONCLUSION AND FUTURE WORK

5.1. Conclusion

The IoT-enabled smart plant disease diagnosis system developed in this project provides an
efficient, scalable, and accurate solution for real-time disease detection in crops. By integrating
IoT devices for real-time data collection with machine learning models, particularly Convolutional
Neural Networks (CNNs), the system offers a significant advancement over traditional methods of
disease detection, which rely heavily on manual observation. The hybrid approach, which
combines CNN-based feature extraction with clustering techniques, achieved an accuracy of
90.12%, demonstrating its superior performance in identifying plant diseases from leaf images.
The system's real-time capabilities provide timely feedback to farmers, allowing them to take
prompt action to mitigate crop loss. Additionally, the design incorporates energy-efficient IoT
devices and cloud-based processing, ensuring that the system can be deployed in rural and remote
areas with limited power and connectivity. Overall, this project demonstrates the potential of IoT
and machine learning technologies to revolutionize agricultural practices and improve crop
management. Table 8 summarizes the key outcomes of the project.

TABLE 8 Summary Of Project

5.2. Future Work

While the system developed in this project offers significant improvements in plant disease
detection, there are several areas where future work could enhance its functionality and
applicability. One potential area of improvement is the integration of more advanced machine
learning techniques, such as ensemble models that combine the strengths of multiple algorithms to
further improve classification accuracy. Another area of focus could be expanding the system’s
ability to handle a wider variety of plant species and diseases. Currently, the system is optimized
for specific crops and diseases, but future iterations should aim to include a broader range of
agricultural products. Additionally, the system's IoT devices could be enhanced to include more
sophisticated sensors that measure environmental conditions, such as temperature, humidity, and
soil moisture. This data could be integrated into the machine learning model to provide more
context for disease diagnosis, further improving the accuracy of predictions. Finally, future work
should explore the development of more user-friendly mobile or web applications that allow
farmers to easily access real-time data and receive actionable insights.

21
REFERENCES
Asif Iqbal, et al. "Role of Internet of Things and Deep Learning Techniques in Plant Disease
Detection and Classification." Sensors, 2023.
Shoaib M, et al. "An advanced deep learning models-based plant disease detection: A review of
recent research." Front Plant Sci, 2023.
Tiwari D, et al. "Potato Leaf Diseases Detection Using Deep Learning." International Conference
on Intelligent Computing and Control Systems (ICICCS), 2020.
Fan Z, Chen Z, Wu J. "Pose Recognition for Dense Vehicles under Complex Street Scenario."
IEEE VCIP 2019, 2019.
Kadam SU, Dhede VM, Khan VN. "Machine Learning Method for Automatic Potato Disease
Detection." International Journal of Recent Research and Review, 2022.
Dhaka VS, et al. "Deep Learning Approaches for Plant Disease Detection: A Review."
Agriculture, 2021.
Wallelign SA, et al. "Tomato Plant Disease Detection Using Deep Learning Techniques." Plant
Methods, 2022.
Zhang K, et al. "Automatic Fruit and Vegetable Disease Detection Using Deep Learning."
Computers and Electronics in Agriculture, 2021.
Ramcharan A, et al. "Deep Learning for Image-Based Cassava Disease Detection." Frontiers in
Plant Science, 2020.
Xie C, et al. "Deep Learning-Based Detection of Rice Diseases Using Image Augmentation
Techniques." IEEE Access, 2020.

22

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy