K-Means Algorithm to Improve Leaf Image Clustering Model for Rice Disease Early Detection

Abstract

This research aims to improve the accuracy of rice leaf image clustering in early disease detection using the K-Means algorithm. The approach used involves the Knowledge Discovery in Databases (KDD) method, which includes data selection, pre-processing, data transformation, data mining, evaluation, and presentation of results. The dataset used consists of images of healthy leaves and leaves infected with diseases such as Bacterial Leaf Blight, Brown Spot, and Leaf Smut. The images are processed through grayscale conversion, noise removal, size adjustment, and data augmentation. The K-Means algorithm is applied to cluster image features based on visual similarity. Evaluation results using Silhouette Score showed that the best clustering was obtained at K=2 with a score of 0.8340, resulting in two main clusters separating healthy and infected images. This study concludes that the K-Means algorithm is able to improve the efficiency and accuracy of rice disease detection, so that it can assist farmers in taking early preventive measures and increase agricultural productivity. This implementation shows significant potential in the development of smart agriculture technology.