<![CDATA[Rice (Oryza sativa) is a strategic Indonesian food commodity that is susceptible to leaf disease attacks, causing decreased productivity and even crop failure. Conventional detection methods based on visual observation have limited accuracy and consistency, so an automated approach based on computer vision technology is needed for more effective early detection. This study applies the K-Nearest Neighbors (KNN) algorithm in rice leaf disease classification using Histogram of Oriented Gradients (HOG) feature extraction. A secondary dataset from Kaggle of 1,400 images covers four categories: Bacterial Leaf Blight, Brown Spot, Leaf Blast, and Healthy. The methodology includes image preprocessing (resize, augmentation, grayscale conversion, normalization), HOG feature extraction, and KNN classification with evaluation on a training-test data ratio of 85:15. The results show that KNN with k=2 achieves optimal performance at a ratio of 85:15 with an accuracy of 90.24%, a precision of 90.27%, a recall of 90.24%, an F1-score of 90.23%, and an efficient computational time of 3.34 seconds. The combination of HOG and KNN is proven to be effective for the automatic classification of rice leaf diseases with high accuracy and good computational efficiency.]]>
Copyrights © 2025
