ML-based approach to potato diseases diagnosis using image processing and whale optimization algorithm for feature selection

In the domain of plant pathology, image processing combined with artificial intelligence has emerged as a crucial approach for accurate plant disease identification. This development has stimulated extensive research involving various methodologies, architectures, and components. A significant challenge in this area is the design and implementation of algorithms capable of handling large datasets efficiently, reducing diagnostic time, and enhancing diagnostic accuracy. Feature selection plays a vital role in this process by eliminating redundant and irrelevant information while retaining the most informative features for disease diagnosis. Nature-inspired feature extraction techniques, such as the Whale Optimization Algorithm (WOA), have gained considerable attention recently. In this study, image processing and machine learning techniques were employed to diagnose Early Blight and Late Blight diseases in potato plants. After extracting texture, color, and shape features using GLCM, Color Moments, and Zernike moments, Support Vector Machine (SVM) was utilized for classification. Additionally, WOA was applied to identify the most significant features contributing to disease diagnosis. The experimental results demonstrated that the proposed method achieved high accuracy rates of 99.04 % for SVM and 96.49 % for WOA. The precision for diagnosing early blight, late blight, and healthy leaves was 96.29 %, 96.29 %, and 97.44 %, respectively. These findings indicate that integrating texture, color, and shape features significantly improves the classification performance of potato leaf diseases and offers valuable insights for disease diagnosis. The proposed approach holds great potential for application in plant disease detection, assisting plant breeders and farmers in early identification of diseases, which can ultimately minimize crop losses and enhance productivity.