Wheat Yellow Rust Disease Infection Type Classification Using Texture Features.

Area of Science:

• Agricultural Science
• Plant Pathology
• Machine Learning

Background:

• Wheat is a vital crop in Pakistan, facing significant yield losses (20-30%) due to rust disease, a major threat to food security.
• Effective management of wheat rust, particularly yellow rust, necessitates precise and early detection of infection types.

Purpose of the Study:

• To propose a machine learning framework for classifying wheat yellow rust infection types.
• To enable early detection and inform remedial measures for minimizing crop yield losses.

Main Methods:

• Collected a dataset of wheat yellow rust infections using mobile cameras.
• Extracted Gray Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP) texture features from grayscale images.
• Employed machine learning classifiers including Decision Tree, Random Forest, LightGBM, XGBoost, and CatBoost for classification.

Main Results:

• The CatBoost model achieved the highest accuracy of 92.30% when using GLCM texture features for classification.
• Evaluated model performance across GLCM, LBP, and combined GLCM-LBP features.

Conclusions:

• The proposed machine learning framework demonstrates potential for accurate wheat yellow rust detection.
• Further improvements in accuracy are achievable by expanding the dataset and incorporating deep learning models.