SVM-Based Detection of Tomato Leaves Diseases

This article introduces an efficient approach to detect and identify unhealthy tomato leaves using image processing technique. The proposed approach consists of three main phases; namely pre-processing, feature extraction, and classification phases. Since the texture characteristic is one of the most important features that describe tomato leaf, the proposed system system uses Gray-Level Co-occurrence Matrix (GLCM) for detecting and identifying tomato leaf state, is it healthy or infected. Support Vector Machine (SVM) algorithm with different kernel functions is used for classification phase. Datasets of total 800 healthy and infected tomato leaves images were used for both training and testing stages. N-fold cross-validation technique is used to evaluate the performance of the presented approach. Experimental results showed that the proposed classification approach has obtained classification accuracy of 99.83%, using linear kernel function.