Pemanfaatan Metode Multiclass Support Vector Machine dalam Klasifikasi Penyakit Daun Kacang Tanah

Abstract

Peanuts are one type of agricultural crop from commodity crops that can provide additional income opportunities for farmers in Indonesia. In addition, the benefits of peanuts are as a source of protein and vegetable fat for human body, so they are also much needed by the food industry. However, in increasing soil productivity there is a decrease in quality and quantity caused by one of the factors, namely plant diseases. Efforts that can be made in maintaining peanut productivity are to prevent early by applying early detection technology. This study presents the application of digital image processing application-based technology using the Multiclass SVM One-Against-One (OAO) strategy to classify the types of leaf disease of peanut plants based on texture feature extraction on the diseased parts of peanut leaves using the Gray Level Co-Occurrence Matrix (GLCM) method. In the classification process using the M-SVM method the OAO strategy will use three kernels, namely polynomial kernel, linear kernel and RBF kernels. Based on the experimental results, the best accuracy is obtained, namely by using GLCM texture feature extraction with a distance of d = 1 and angle 90 degree of and classified using the M-SVM method, the OAO strategy with polynomial kernels provides the highest accuracy results, namely 96.39% for leaf spot class, 92.79% for leaf rust class, 96.39% for eye spot class and 100% for normal class