IMAGE CLASSIFICATION OF POTATO LEAF DISEASES USING K-NEAREST NEIGHBOR WITH ADDITIONAL CHLOROSIS AND NECROSIS FEATURES
Abstract
This study aims to improve the accuracy of potato leaf disease classification by addressing the limitations of existing models that rely solely on general features such as shape, texture, and color. Potato leaf diseases such as early blight and late blight can significantly reduce crop yields if not detected early and accurately. However, many current models overlook critical biological symptoms like chlorosis and necrosis, which impacts detection performance. To overcome this issue, this study proposes a classification approach using the K-Nearest Neighbor (KNN) algorithm, integrating chlorosis and necrosis features alongside conventional features. The dataset consists of three classes Potato Early blight, Potato Late blight, and Potato healthy, sourced from Kaggle. Feature extraction is conducted using color histograms, shape descriptors, texture analysis, and binary masks for chlorosis and necrosis regions. The model is evaluated using an 80:20 train-test split and assessed using a confusion matrix, accuracy, precision, recall, and F1-score. The results show that the addition of chlorosis and necrosis features significantly enhances classification performance, with the best accuracy reaching 90%. This demonstrates that the proposed method effectively improves the KNN algorithm’s ability to distinguish between disease types and healthy leaves, thereby addressing the initial problem of limited classification accuracy.