Corn Leaf Diseases Classification Using CNN with GLCM, HSV, and L*a*b* Features

Abstract

Corn leaf diseases can damage plants and reduce crop yields, thus affecting the quality and quantity of corn production. This study aims to classify corn leaf diseases using the Convolutional Neural Network (CNN) method with different color features, namely Gray Level Co-Occurrence Matrix (GLCM), HSV, and L*a*b*. The dataset consists of 1,739 corn leaf images, which are divided into four disease classes: Blight, Common Rust, Gray Spot, and Healthy. The data is split into training and testing sets using an 80:20 ratio. Two testing scenarios were conducted: individual feature evaluation and feature combination. The results show that in the first scenario, the L*a*b* feature provides the best accuracy at 91.75%, followed by the HSV feature with an accuracy of 90.29%, and GLCM with an accuracy of 78.40%. In the second scenario, the combination of HSV and L*a*b* features results in the highest accuracy of 92.48%, indicating that combining color and brightness information can improve the model's performance. The combination of GLCM and L*a*b* features results in an accuracy of 91.75%, while the combination of GLCM and HSV results in an accuracy of 90.29%. These findings demonstrate that integrating HSV and L*a*b features enhances CNN performance in corn leaf disease classification, outperforming individual feature- based approaches, thus contributing to more effective AI-based agricultural disease diagnosis.