Enhancing image based classification for crop disease detection using a multiclass SVM approach with kernel...

Area of Science:

• Agricultural Science
• Computer Science
• Machine Learning

Background:

• Plant diseases significantly impact agricultural production, threatening food security and economic stability.
• Specific diseases like yellow rust and anthracnose cause substantial yield losses in crops such as wheat, cotton, and mango.

Purpose of the Study:

• To develop and evaluate a machine learning framework for early and precise detection of diverse crop leaf diseases.
• To compare the effectiveness of different multiclass Support Vector Machine (SVM) kernels for disease classification.

Main Methods:

• A machine learning pipeline involving image preprocessing, segmentation (GraphCut), texture-based feature extraction, and classification was implemented.
• A dataset of 9,111 augmented images across multiple crops was used for model training and validation.
• Stratified 5-fold cross-validation was employed to systematically evaluate SVM kernel performance.

Main Results:

• The linear kernel SVM demonstrated superior performance, achieving 99.0% accuracy, 98.6% precision, 98.7% recall, and 98.6% F1-score.
• The proposed method, combining bilateral filtering, GraphCut segmentation, and texture features, outperformed previous SVM-based approaches.

Conclusions:

• Kernel selection and preprocessing significantly enhance the accuracy of plant disease classification.
• The findings support the development of scalable and reliable automated plant disease detection systems, with potential for future deep learning comparisons.