Automated detection of rice plant diseases using dual stage thresholding and twin support vector machine

Abstract

Rice plants are susceptible to various diseases such as brown spot, BLB, and blast, caused by viral, bacterial, or fungal infections, which significantly affect both the quantity and quality of rice production. This study introduces an automated method for detecting these diseases using dual thresholding (DT) in segmentation combined with twin support vector machine (TW-SVM) classification. Early detection and accurate identification of rice leaf diseases are crucial for effective management and optimization of production. The proposed method leverages the strengths of TW-SVM, including its ability to handle high-dimensional data efficiently. The approach is compared with three SVM-based techniques: basic SVM, least-square SVM, and proximal SVM. Simulations are performed using images from both a public dataset and a real-time drone image dataset. Thirteen features, including color, texture, and shape, are extracted for classification. Results show that the proposed dual stage thresholding (DST) TW-SVM achieves superior performance in terms of time complexity and accuracy, with 95% accuracy on the public dataset and 99.3% accuracy on the drone image dataset.