Cascading LBP-GLCM-JST Untuk Model Klasifikasi Makroskopis Kayu Komersial

Abstract

Manual classification of wood species is time-consuming, consequently, a faster method is needed. One method that can be used is image processing. Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) methods have been widely used for feature extraction but show varying accuracy results. Research is needed to improve the accuracy of image processing in terms of feature extraction. LBP and GLCM are algorithms for feature extraction for images. The classification process used a Backpropagation Artificial Neural Network (ANN BP), which can distinguish several types of wood. This study compared the combined methods of LBP, GLCM, and GLCM for image feature extraction. The best accuracy obtained was 98% for the combined LBP and GLCM methods, while the best accuracy for the GLCM method was 86%. The best JSTPB model for the combined method of LBP and GLCM was parameter 1 hidden layer, 10 hidden layer neurons, learning rate 0.1, and momentum 0.1. The best JSTPB model for the GLCM method was parameter 1 hidden layer, 20 hidden layer neurons, learning rate 0.5, and momentum 0.5. From a total of 80 parameter variations, the accuracy of the LBP and GLCM method was always higher than the GLCM method.