Optimizing Rice Disease Classification: A Comparative Study of Different Color, Shape and Texture Features

Rice holds significant importance as a staple food across Asia. Despite its crucial role, the rice industry grapples with challenges in disease classification. Current approaches utilizing ANN and SVM often rely on pre-processing and segmentation of the diseased portion of the rice plant before classification, which prevents automation. This study presents an automated, more efficient method by eliminating pre-processing and segmentation steps as well as utilizing the whole image. Leveraging features such as color histogram, GLCM, color moments, Zernike moments, and Legendre moments, the study trains SVM and ANN models—recognized for their effectiveness—to classify thirteen rice diseases, achieving accuracies of 86.23% and 83.93% respectively.