Below is a well-structured Chapter Three (Methodology) for your thesis titled “Computer Vision
Technique for Disease Detection in Ginger Using Deep Learning (CNN)”, written according to your three
specific objectives, and including in-text citations, diagrams (ASCII-style), and references.
---
**CHAPTER THREE
METHODOLOGY**
3.1 Introduction
This chapter describes the methodological framework employed in developing a computer vision system
for detecting diseases in ginger plants using deep learning. The methodology covers the system design,
data acquisition, preprocessing, model development, interface design, and system evaluation. This
approach is guided by the study objectives, which include designing a user-friendly interface for real-
time detection, automating the disease identification process, and employing deep learning algorithms
for accurate classification of healthy and diseased ginger plants.
---
3.2 Research Design
The study adopts an experimental research design supported by computational modelling. The process
involves collecting ginger leaf images, preprocessing them, training a Convolutional Neural Network
(CNN) model, and integrating the model into a real-time interface for farmers. The overall workflow is
shown in Figure 3.1.
Figure 3.1: System Workflow Diagram
+----------------------+
| Image Acquisition |
+----------+-----------+
+----------------------+
| Image Preprocessing|
| (Resize, Augment etc)|
+----------+-----------+
+----------------------+
| CNN Model Training |
+----------+-----------+
+----------------------+
| Disease Classification|
+----------+-----------+
+----------------------+
| Real-time Interface |
+----------------------+
---
3.3 Data Collection
Images of healthy and diseased ginger plants were sourced from field surveys, agricultural extension
agencies, and publicly available datasets. The dataset includes common ginger diseases such as leaf
spot, bacterial wilt, and fungal blight. Each image was labeled by experts to ensure reliability, consistent
with the practice in computer vision-based plant pathology studies (Ferentinos, 2018).
---
3.4 Data Preprocessing
Collected images undergo preprocessing to enhance model performance:
Resizing: All images were resized to pixels to match CNN input requirements.
Normalization: Pixel intensity values were normalized to a range of 0–1.
Data Augmentation: Rotation, flipping, zooming, and brightness adjustments were applied to increase
dataset diversity and mitigate overfitting (Shorten & Khoshgoftaar, 2019).
Segmentation: Background noise was reduced using threshold-based segmentation to keep only leaf
regions.
---
3.5 System Design
The system is designed according to the three research objectives.
3.5.1 User-Friendly Real-Time Interface Design (Objective 1)
A simple, intuitive, mobile-friendly interface was designed to allow farmers to:
Capture real-time images of ginger leaves
View instant classification results (healthy or specific disease)
Access recommendations for disease management
The interface was implemented using Flask/TensorFlow Lite Web API for mobile deployment. The design
principles follow usability standards proposed in agricultural decision-support systems (Nikas et al.,
2023).
Figure 3.2: Interface Design Structure
+---------------------------------------------------------+
| Real-Time Detection App |
+---------------------------------------------------------+
| [Camera Capture] [Upload Image] |
| |
| Display Area: |
| --------------------------- |
| | Input Leaf Image | Result: Healthy/Diseased | |
| --------------------------- |
| |
+---------------------------------------------------------+
---
3.5.2 Automated Disease Identification Process (Objective 2)
To reduce the time and effort required in disease detection, the system automates:
1. Image acquisition
2. Preprocessing
3. Feature extraction using CNN filters
4. Classification
5. Recommendation generation
Automation minimizes manual inspection, which is traditionally slow and prone to human error.
According to Too et al. (2019), automating plant disease detection significantly accelerates agricultural
diagnostics.
---
3.5.3 Deep Learning Algorithm for Classification (Objective 3)
The classification model is based on a Convolutional Neural Network (CNN) due to its effectiveness in
plant disease detection tasks (Mohanty et al., 2016).
CNN Architecture Used
Input (224x224x3)
|
Conv Layer (32 filters, 3x3)
ReLU Activation
MaxPooling (2x2)
Conv Layer (64 filters)
ReLU + MaxPooling
Flatten
Fully Connected Layer (128 units)
Dropout (0.5)
Output Layer (Softmax: Healthy/Diseased)
The model was trained using categorical cross-entropy loss, Adam optimizer, and batch size of 32, with
early stopping to prevent overfitting.
---
3.6 Model Training and Validation
The dataset was split into:
70% training
20% validation
10% testing
Performance metrics included accuracy, precision, recall, and F1-score. Training was done on a GPU-
enabled system to accelerate computation.
---
3.7 System Implementation
The trained model was serialized and integrated into the interface for real-time inference. The
application supports:
Mobile camera input
Offline prediction using TensorFlow Lite
Cloud-based prediction (optional)
---
3.8 Evaluation Metrics
System accuracy and usability were evaluated through:
Technical evaluation (accuracy, confusion matrix)
User evaluation (farmer usability testing)
These metrics ensure the system is both effective and practical for real-world deployment.
---
3.9 Summary
This chapter presented the methodological approach used to design a computer vision-based system for
ginger disease detection. It covered dataset preparation, model development, interface
implementation, and evaluation strategies aligned with the study objectives.
---
REFERENCES
Ferentinos, K.P. (2018). Deep learning models for plant disease detection and diagnosis. Computers and
Electronics in Agriculture, 145, 311–318.
Shorten, C., & Khoshgoftaar, T. (2019). A survey on image data augmentation for deep learning. Journal
of Big Data, 6(60), 1–48.
Too, E.C., et al. (2019). A comparative study of fine-tuning deep learning models for plant disease
identification. Computers and Electronics in Agriculture, 161, 272–279.
Mohanty, S.P., Hughes, D.P., & Salathe, M. (2016). Using deep learning for image-based plant disease
detection. Frontiers in Plant Science, 7, 1419.
Nikas, A., et al. (2023). Usability principles in digital agricultural decision support systems. Agricultural
Systems, 207, 103620.
---
If you want, I can also provide:
✅ Chapter Four (System Implementation & Results)