MEKELLE INSTITUTE OF TECHNOLOGY

DEPARTMENT: ELECTRONICS AND COMMUNICATION ENGINEERING

PROJECT TITLE: PLANT DISEASE DETECTION USING MACHINE

LEARNING IMPLEMENTED WITH A SELF-DRIVING ROBOT

MEMBERS ID NUMBER Ph.NO Email

HAGAZI DESTA MIT/UR/105/11 0987792434 [email protected]

KIFLU TEKLAY MIT/UR/131/11 0927781230 [email protected]

H/MIKIAL H/MARIAM MIT/UR/273/11 0996172715 [email protected]

Supervisor Name: AWET KIDANE

Submission date: 15/06/2017 E.C

 TABLE OF CONTENTS
Chapter one................................................................................................................................................ 1
introduction................................................................................................................................................ 1
1.1 Background................................................................................................................................... 1
1.2 Problem Statement.......................................................................................................................2
1.3 Research Questions...................................................................................................................... 2
1.4 Objectives..................................................................................................................................... 2
1.5 Significance of the Study...............................................................................................................2
1.6 Scope and Limitations...................................................................................................................3
Chapter two................................................................................................................................................ 3
Literature Review........................................................................................................................................3
2.1 Machine Learning in Agriculture...................................................................................................3
2.2 Robotics in Agriculture..................................................................................................................4
2.3 Image Processing for Plant Disease Detection..............................................................................4
2.4 Summary of Literature Findings....................................................................................................5
2.5 Summary of Related Studies.........................................................................................................5
2.6 Research Gaps...............................................................................................................................5
Chapter three............................................................................................................................................. 6
Methodology.............................................................................................................................................. 6
3.1 Research Design............................................................................................................................6
3.2 Data Collection Methods and Sampling Techniques.................................................................... 7
3.3 Data Analysis Techniques..............................................................................................................7
Chapter four................................................................................................................................................7
Expected Outcomes....................................................................................................................................7
Chapter five................................................................................................................................................ 8
Timeline...................................................................................................................................................... 8
Chapter six.................................................................................................................................................. 8
Resources and Budget................................................................................................................................ 8
References.................................................................................................................................................. 8

I
Chapter one

introduction

1.1 Background

Plant diseases are a major threat to global food security, leading to significant losses
in agricultural productivity each year. The economic impact of these diseases is
severe, especially for smallholder farmers who rely on agricultural yields for their
livelihoods. Early detection and accurate diagnosis of plant diseases are crucial to
mitigating these losses and ensuring food security. Traditional methods for disease
detection involve visual inspection by farmers or agricultural experts, which can be
subjective, time-consuming, and prone to human error (Mahlein, 2016). Furthermore,
these methods require extensive field surveys, which may not always be feasible for
large-scale farms.

Recent advancements in machine learning (ML) and robotics have opened new
avenues for automating plant disease detection, offering efficient and scalable
solutions. Machine learning algorithms, particularly deep learning models, have
demonstrated high accuracy in detecting and classifying plant diseases based on leaf
symptoms captured in images . When integrated with self-driving robots , these
technologies can enable real-time monitoring of large agricultural fields with minimal
human intervention. By utilizing computer vision and sensor-based data collection,
these autonomous systems can detect early signs of infection, allowing farmers to take
timely preventive measures. This reduces the overuse of chemical pesticides and
promotes sustainable agricultural practices .

The integration of ML with self-driving robots for plant disease detection is not only a
technological advancement but also a transformative approach to modern farming.
These intelligent systems can analyze vast amounts of agricultural data, improve
precision in disease diagnosis, and enhance decision-making processes for farmers.
Additionally, such innovations align with the goals of precision agriculture, where
resources such as water, fertilizers, and pesticides are optimized based on data-driven
insights . As research in this field progresses, the development of more robust and
adaptable models will further enhance the effectiveness of automated plant disease
detection systems, contributing to increased crop yields and overall agricultural
sustainability.

1
1.2 Problem Statement

The current methods of plant disease detection are inefficient, especially for large-
scale agricultural operations. Manual inspection is prone to human error and is not
scalable. There is a need for an automated system that can accurately detect plant
diseases in real-time, allowing for timely intervention and reducing crop losses. This
project aims to address this gap by developing a machine learning-based system
implemented on a self-driving robot for autonomous plant disease detection.

1.3 Research Questions

 How can machine learning algorithms be effectively used to detect plant diseases
from images of crops?
 What is the optimal design for a self-driving robot capable of navigating
agricultural fields and capturing high-quality images of plants?
 How can the integration of machine learning and robotics improve the accuracy
and efficiency of plant disease detection compared to traditional methods?

1.4 Objectives

 To develop a machine learning model capable of accurately detecting plant

diseases from images.
 To design and implement a self-driving robot that can autonomously navigate
agricultural fields and capture images of plants.
 To integrate the machine learning model with the self-driving robot for disease
detection.
 To evaluate the performance of the system in terms of accuracy, efficiency, and
scalability.

1.5 Significance of the Study

This study has the potential to revolutionize the way plant diseases are detected and
managed in agriculture. By automating the detection process, the system can
significantly reduce the time and labor required for disease identification, allowing
farmers to take timely action to prevent crop losses. The integration of machine
learning and robotics can also provide a scalable solution that can be adapted to
different crops and environments. This research will contribute to the fields of
precision agriculture, robotics, and machine learning, and has the potential to improve
food security and sustainability.

2
1.6 Scope and Limitations

The scope of this study is limited to the detection of common plant diseases using
image-based machine learning models. The self-driving robot will be designed to
operate in controlled agricultural environments, such as greenhouses or small farms.
The system will be tested on a limited number of crop types and diseases.

The limitations include the need for high-quality image data for training the machine
learning model and the potential challenges in navigating complex agricultural
terrains. Additionally, the system may face constraints due to the limited processing
capacity of the device used, storage limitations, and the availability and reliability of a
wireless camera for capturing plant images.

Chapter two

Literature Review

The integration of advanced technologies such as machine learning, robotics, and

image processing has significantly transformed modern agriculture. These innovations
enhance precision farming, automate labor-intensive tasks, and improve the early
detection of plant diseases. This section reviews the existing literature on machine
learning in agriculture, robotics in agriculture, and image processing for plant disease
detection while highlighting relevant studies and findings.

2.1 Machine Learning in Agriculture

Machine learning (ML) has become a powerful tool in precision agriculture,

particularly for plant disease detection. Traditional methods of disease identification
rely on expert knowledge and manual inspection, which can be time-consuming,
labor-intensive, and prone to human error. With advancements in artificial
intelligence, ML algorithms can analyze large datasets to recognize patterns in plant
health and detect diseases at an early stage (Zhang et al., 2020).

Among various ML models, Convolutional Neural Networks (CNNs) have

demonstrated superior performance in image-based plant disease detection. CNNs
automatically extract relevant features such as leaf texture, color, and shape, allowing
for accurate classification of plant diseases [8]. Several studies have shown that CNN-
based models outperform traditional image-processing techniques by achieving high
classification accuracy, often exceeding 90% in specific datasets [4]. Other deep
learning models, such as Recurrent Neural Networks (RNNs) and Transformers, are
being explored for agricultural applications, particularly for analyzing time-series data
and optimizing crop health prediction [5].

However, challenges persist in implementing ML for plant disease detection. The

availability of high-quality, labeled datasets is a significant constraint, as variations in

3
lighting, environmental conditions, and camera quality can impact model accuracy .
Additionally, deploying deep learning models in real-time agricultural settings is
computationally demanding, making it difficult to implement on low-power edge
devices [6].

2.2 Robotics in Agriculture

The application of robotics in agriculture has expanded due to the demand for
automation in tasks such as crop monitoring, harvesting, and disease detection.
Autonomous agricultural robots can enhance efficiency by reducing dependency on
manual labor and enabling continuous field monitoring [2]. These robots use a
combination of machine vision, LiDAR, GPS, and deep learning algorithms to
navigate agricultural environments, identify diseased crops, and take corrective
actions such as targeted spraying or early disease intervention [3].

Autonomous self-driving robots are increasingly used for precision farming. Path
planning algorithms such as A and Dijkstra’s algorithm help these robots move
efficiently while avoiding obstacles . In disease detection applications, robots
equipped with high-resolution cameras and multispectral sensors capture detailed
images of crops, which are then processed using ML-based classification models .

Despite these advancements, several challenges hinder the widespread adoption of

agricultural robots. Agricultural terrains can be highly uneven, making autonomous
navigation complex. Additionally, power consumption and battery life remain key
concerns, as these robots require sustainable energy sources for prolonged field
operations [9].

2.3 Image Processing for Plant Disease Detection

Image processing plays a crucial role in automated plant disease detection by enabling
the extraction and analysis of visual features related to plant health. Traditional image
processing techniques, such as thresholding, edge detection, and color-based
segmentation, have been widely used for disease classification, but their effectiveness
is often limited by variations in environmental conditions .

Modern approaches integrate computer vision and deep learning to enhance accuracy.
Preprocessing techniques such as histogram equalization, noise reduction, and
background subtraction help improve image quality before analysis . CNN-based
models further refine these images by extracting hierarchical features, allowing for
more precise classification of plant diseases [1].

Recent advancements in hyperspectral imaging and thermal imaging have further

improved disease detection by capturing detailed spectral signatures of plants. These
techniques allow early identification of stress factors that may not be visible to the
human eye [10]. However, implementing such advanced imaging solutions requires
specialized hardware, which can be costly and may not be easily accessible for small-
scale farmers [7].

4
2.4 Summary of Literature Findings

The review of existing literature highlights significant progress in applying machine

learning, robotics, and image processing to agriculture. While CNN-based models
have demonstrated effectiveness in plant disease detection, challenges related to
dataset availability, environmental variations, and computational constraints persist.
Robotics has introduced automation in agricultural monitoring, but mobility
limitations and power constraints require further optimization. Image processing
techniques have improved disease detection accuracy, but the integration of advanced
imaging technologies necessitates further research to enhance affordability and
accessibility.

By addressing these challenges, future research can develop more robust and scalable
solutions for precision agriculture, ultimately improving crop health monitoring and
early disease detection.

2.5 Summary of Related Studies

Several studies have explored the use of machine learning for plant disease detection.
For example, [5] developed a CNN-based model for detecting diseases in tomato
plants with high accuracy. Similarly, [9] proposed a robotic system for autonomous
crop monitoring, but the integration of machine learning for disease detection remains
underexplored.

2.6 Research Gaps

While there have been significant advancements in machine learning and robotics for
agriculture, there is a lack of integrated systems that combine both technologies for
plant disease detection. Additionally, most existing systems are tested in controlled
environments, and their scalability to larger, more complex agricultural settings is
unclear.

5
Chapter three

Methodology

3.1 Research Design

This study will adopt a mixed-methods research approach, integrating both

quantitative and qualitative methodologies to ensure a comprehensive evaluation of
the self-driving robot's performance in plant disease detection. The combination of
these approaches allows for a more robust analysis, capturing both numerical data
related to system accuracy and qualitative insights from users and field testing.

On the quantitative side, the research will focus on evaluating the performance of the
machine learning (ML) model used for plant disease detection. Key metrics such as
accuracy, precision and computational efficiency will be analyzed to assess the
reliability of the model. The ML model will be trained on a dataset of labeled plant
disease images and tested in real-world agricultural environments, such as
greenhouses and small farms. Additionally, the robot’s navigation performance will
be quantitatively assessed using metrics like path accuracy, obstacle avoidance
success rate, and real-time processing speed to determine its efficiency in controlled
agricultural settings.

On the qualitative side, the study will collect user feedback from researchers to
evaluate the system’s usability, ease of operation, and practical applicability.
Structured interviews and surveys will be conducted to gather insights into the robot’s
effectiveness, user satisfaction, and potential improvements. Observational studies
will also be employed to examine how the robot interacts with the environment,
particularly in challenging conditions such as uneven terrain, varying lighting
conditions, and dense vegetation.

By integrating both quantitative performance metrics and qualitative user evaluations,

this research aims to provide a well-rounded analysis of the self-driving robot's
capabilities. The mixed-methods approach will also help in identifying key areas for
improvement, ensuring that the technology is not only technically efficient but also
practically useful for end-users in the agricultural sector.

6
3.2 Data Collection Methods and Sampling Techniques

 Image Data Collection: High-resolution images of healthy and diseased plants

will be collected from agricultural fields and publicly available datasets.
 Robot Navigation Data: The self-driving robot will be tested in controlled
agricultural environments to collect data on navigation and image capture.
 Sampling: The dataset will include a diverse range of crop types and diseases to
ensure the model's generalizability.

3.3 Data Analysis Techniques

 Machine Learning Model Development: machine learning model will be trained

on the collected image data to detect plant diseases.
 Robot Navigation Analysis: The robot's navigation performance will be evaluated
based on accuracy, speed, and obstacle avoidance.
 System Integration: The machine learning model will be integrated with the
robot's software for disease detection.

7
Chapter four

Expected Outcomes

The expected outcomes of this research include:

 A highly accurate machine learning model for plant disease detection.

 A functional self-driving robot capable of autonomously navigating agricultural
fields and capturing images.
 An integrated system that can detect plant diseases providing actionable insights
to farmers.
 Evaluation results demonstrating the system's accuracy, efficiency, and scalability.

8
Chapter five Timeline

Task Start Date End Date

Literature Review 13/06/2017 E.C 30/06/2017 E.C
Data Collection 01/07/2017 E.C 07/07/2017 E.C
MachineLearning Model Development 08/07/2017 E.C 24/07/2017 E.C
Robot Design and Implementation 25/07/2017 E.C 24/08/2017 E.C
System Integration 25/08/2017 E.C 24/09/2017 E.C
Testing and Evaluation 25/09/2017 E.C 30/09/2017 E.C
Thesis Writing and Submission 01/10/2017 E.C 10/10/2017 E.C

Chapter six

Resources and Budget

 Hardware: Self-driving robot components (e.g., cameras, sensors, motors etc …),
high processing computing equipment (e.g., GPUs for model training).
 Software: Machine learning frameworks (e.g., TensorFlow, PyTorch), robot
operating system (ROS) car kit.
 Datasets: Publicly available plant disease datasets, custom dataset collection.
 Budget: Estimated total cost of 200,000 ETB, including hardware, software, and
miscellaneous expenses.

References

1. Amara, J., Bouaziz, B., & Algergawy, A. (2017). A deep learning-based approach
for banana leaf diseases classification. Lecture Notes in Informatics, 1-7.
2. Bac, C. W., Van Henten, E. J., Hemming, J., & Edan, Y. (2014). Harvesting
robots for high-value crops: State-of-the-art review and challenges ahead. Journal
of Field Robotics, 31(6), 888-911.
3. Bechar, A., & Vigneault, C. (2016). Agricultural robots for field operations:
Concepts and components. Biosystems Engineering, 149, 94-111.
4. Ferentinos, K. P. (2018). Deep learning models for plant disease detection and
diagnosis. Computers and Electronics in Agriculture, 145, 311-318.
5. Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A
survey. Computers and Electronics in Agriculture, 147, 70-90.
6. Liakos, K. G., Busato, P., Moshou, D., Pearson, S., & Bochtis, D. (2018).
Machine learning in agriculture: A review. Sensors, 18(8), 2674.
7. Mahlein, A. K., Rumpf, T., Welke, P., Dehne, H. W., Plümer, L., Steiner, U., &
Oerke, E. C. (2012). Development of spectral indices for detecting and
identifying plant diseases. Remote Sensing of Environment, 128, 21-30.

9
8. Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using deep learning for
image-based plant disease detection. Frontiers in Plant Science, 7, 1419.
9. Oberti, R., Marchi, M., Tirelli, P., Calcante, A., Iriti, M., & Borghese, A. N.
(2016). Automatic detection of powdery mildew on grapevine leaves by image
analysis: Optimal view angle range to increase sensitivity. Computers and
Electronics in Agriculture, 127, 155-165.
10. Rumpf, T., Mahlein, A. K., Steiner, U., Oerke, E. C., Dehne, H. W., & Plümer, L.
(2010). Early detection and classification of plant diseases with support vector
machines based on hyperspectral reflectance. Computers and Electronics in
Agriculture, 74(1), 91-99.

10