Implementasi Metode Gray Level Co-occurrence Matrix dalam Identifikasi Jenis Daun Tengkawang

Abstract

Tengkawang or known as Borneo tallow nut is now difficult to find due to unsustainable forestry practices and high levels of forest destruction. The method used in this research was pattern recognition. The process of identifying Tengkawang plants was carried out through this process: image acquisition, image cutting and background removal of tengkawang leaf image, RGB image conversion of tengkawang leaf into gray scale, threshold limit determination with certain value, feature extraction with GLCM method (spatial distance 1, 2, and 3 pixels), and morphology, so the pattern of Tengkawang leaf image could be obtained. The image pattern was classified using Back Propagation Neural Network algorithm. The output is a software for identifying the types of tengkawang leaf. The results of identification testing of non-tengkawang leaf species show that using a total of 16 random samples of test images, an accuracy of 87.5% is obtained. The identification rate of tengkawang leaf image with spatial distance of 1, 2, and 3 pixels from total 24 random sample of test image shows 100% accuracy level. Training with 2 pixel spatial spacing has the lowest iteration, i.e. 10 iterations. The result of identifying damaged tengkawang leaf image on the edge has an effect on the extraction of morphological characteristics.