Integrating Color Segmentation and Texture Analysis for Citrus Leaf Disease Identification through K-Means and Local Binary Pattern

Abstract

Early identification of citrus leaf diseases is important for reducing crop losses and supporting effective disease management. This study proposes a citrus leaf disease identification framework combining K-means color segmentation and Local Binary Pattern (LBP) texture analysis. The method was evaluated using 90 citrus leaf images from three disease classes: Citrus Canker, Leaf Miner, and Sooty Mold. Images were preprocessed through cropping, resizing, and color space transformation. K-means clustering in the Lab color space was used to segment diseased regions, while LBP extracted texture features. Classification was performed using Euclidean Distance and evaluated with 10-fold cross-validation. The K-means-based method achieved an average accuracy of 76.65%, while LBP alone obtained 45.65%. Combining K-means and LBP produced the best performance, with an average accuracy of 80.02%. These results indicate that integrating color and texture features provides a more effective representation of citrus leaf disease symptoms and improves classification performance.