Tomato leaf diseases classification using image processing and weighted ensemble learning

Abstract

AbstractIn ensemble methods, multiple base classifiers with different performances are used to increase classification accuracy. This study proposes a novel weighted majority voting ensemble approach to classify red green blue images to identify the bacterial spot, late blight, leaf mold, sectorial leaf spot, target spot, early blight, and healthy tomato leaves. Color, textural, and shape features of the images were extracted to be used as inputs for base classifiers, followed by selecting the effective features using the relief method to increase the prediction performance. Six machine learning methods, that is, support vector machine, decision tree, random forest, k‐nearest neighbors, naïve Bayes, and discriminative analysis, were used as base classifiers, which resulted in 89.81%, 79.81%, 91.53%, 85.91%, 44.42%, and 82.84% accuracies, respectively. Then, simple majority and weighted majority voting ensemble methods were utilized to improve disease classification, the accuracies of which were 93.49% and 95.58%, respectively. The performance of the proposed method was compared with that of two well‐known deep learning methods, namely GoogLeNet and AlexNet, which exerted poor results. The findings of this study showed that the proposed framework based on weighted majority voting outperformed the base machine learning models in classifying tomato diseases.