Identifikasi Varietas Durian Berdasarkan Citra Daun Menggunakan K-Nearest Neighbor dengan Ekstraksi Tekstur Co-occurrence Matrix

Abstract

Durian (Durio zibethinus Murray) is the name of tropical plant from Southeast Asia. The diversity of durian and the physical similarities between them have caused difficulty to identify durian. This research developed a system to identify durian by using co-occurrence matrix as the feature extraction and K-NN as the classifier based on durian leaf texture. This research used 9 variety of durians and each variety has 10 leaf images. The result of this research showed that the cropping of images and selection of texture features can improve the performance identification. The best accuracy in this research was 74.44% that was obtained when K = 5, angle = 450, and distance = 2 pixels. This accuracy was achieved by using five leaf texture features, namely contrast, entropy, information measures of correlation 1, sum average, and sum entropy.