The Implementation of K-Means Algorithm as Image Segmenting Method in Identifying the Citrus Leaves Disease

Abstract

The purpose of this study is to identify the disease on citrus leaves. A digital imagery makes it possible to identify disease automatically. Three diseases to be identified in this research are CVPD (Citrus Vein Phloem Degeneration), Downy Mildew, and Cendawan Jelaga. The research will study the implementation of an image segmentation to analyze the citrus leaves diseases. The method that will be used to do image segmentation is K-Means. The segmentation which will be carried out consist of two kinds, namely a leaf segmentation and a disease segmentation. After segmentation process, the results of disease segmentation are classified by using the K-Nearest Neighbor (K-NN) algorithm to know its disease class. From data analysis, the results of the optimal cluster shows that the leaf segmentation consist of 2 clusters and the disease segmentation consist of 9 clusters. While the obtained optimal parameter K gives score of 4. The accuracy percentage for disease identification in this study is 90.83%. Furthermore, the analysis states that the accuracy can be more increased by using a minimum bound parameters. Finally, overall the results show the optimal value at the minimum bound of 3%, its accuracy can be increased to 99.17%.