Implementation of Fuzzy K-Nearest Neighbor (K-NN) Algorithm to Identify Grape Plant Diseases

Abstract

One of the parts of the vine that can be attacked by disease is the leaf. There are 5 types of diseases that can attack grape leaves Black Rot, Leaf Blight (Pseudocercospora Vitis), Black Measles (Phaeomoniellaaleophilum, Phaeomoniella chlamydospora), Powdery Mildew (Ersyphe Necator Burr), and Downy Mildew (Plasmopara Viticola). The symptoms caused by each disease will show different colors and textures of spots on the leaves. For this reason,a system is needed toidentify the type of disease that attacks grape leaves so that appropriate control can be carried out.In this research, a system is built to identify grape leaf disease using a method that is able to recognize the texture on the leaves. The GLCM method is one of the methods for texture extraction in images. In the case of grape disease identification, the selection of color extraction methods is needed in order to achieve good results in the system. One method that is quite good at extracting color features is HSV because the colors in the HSV model are the same as the colors captured by human senses and are able to separate the intensity components of color images. After GLCM and HSV, then theclassification process using the FKNN method by combining Fuzzy and KNN Classifier techniques. The FKNN method has two advantages, where this algorithm is able to consider the ambiguous nature of neighbors, and provide strength in the intances that are in a class, so that the classification process can be done moreobjectively.The results of testing 2720 images as training data with 200 images as testing data show the accuracy value obtained is 92.5%.