Color and Texture Feature Extraction for Disease Identification in Chili Leaves Using K-Nearest Neighbors

Manual identification of chili leaf diseases has the weakness of subjectivity, which impacts the decline in harvest productivity. This study aims to build an accurate automatic classification system using a machine learning approach. The research methodology integrates the extraction of Hue, Saturation, Value (HSV) color features and Gray Level Co-occurrence Matrix (GLCM) texture on a dataset of 1,856 images divided with a ratio of 80:20. Hyperparameter optimization was performed using Grid Search on the K-Nearest Neighbors (K-NN) algorithm to find the best performance. The test results show that the optimal configuration is achieved at a value of K = 3 with the Manhattan distance metric, which produces a test accuracy of 92%. It is concluded that the integration of color and texture features with appropriate parameter optimization is proven to be effective as a reliable and efficient diagnostic solution.