Identifikasi Varietas Durian Berdasarkan Tekstur Daun dengan Metode Ekstraksi GLCM Menggunakan PNN

Abstract

The diversity and similarity of durian varieties have caused some difficulties to identify durian. In this research, the identification is performed by durian leaf texture extraction. The leaf is selected because it is a characteristic that can be used to differ the type of plant. Identification is performed by texture extraction to anticipate the difficulty of morphology based identification, since durian varieties have similar leaves shape. In this research, identification based on leaf texture is performed using GLCM and PNN. This experiment is based on 0o, 45o, 90o, and 135o GLCM angle orientation. Each angles are tested at 1 pixels, 2 pixels, and 3 pixels distances. They are also tested for PNN smoothing parameter (σ) ranges from 0.1 to 9. Identification is performed using 90 images from 9 varieties of durian and the dimension of image is 281 x 281 pixel. The highest accuracy achieved is 76.67% in 0o angle experiment with 1 pixels distance and 0.1 smoothing parameter (σ).