Wood Type Identification Using Support Vector Machine Base on Image Data

Abstract

The wood type identification can be studied as a science, but the ability to identify the type of wood can only be obtained through a long training process, repeatedly and continuously. The problems that appear when workers are not skilled enough to identify the type of wood are the longer time of business process, and thus certainly will added more cost. The wood type identification system is very necessary to provide an alternative solution for this problem. The wood samples had obtained from Balai Pemantauan Pemanfaatan Hutan Produksi Wilayah XII Palangka Raya, which is consists of 4 types of wood, namely: Johar, Jati, Rasamala and Sengon wood. The image of wood type data has obtained using a microscopic camera, and each type has 24 images. Total data in this study are 96 images with has an initial size 480 x 640 pixels and converted into 96 x 128 pixels size. This RGB sample data image is converted into a grayscale images. The image processing in this study using Two - Dimensional Principal Component Analysis (2D-PCA) feature extraction method, which is serves to reduce the images dimensions without removing the important information. The Support Vector Machine (SVM) with kernel function Radial Basis Function (RBF) and polynomial kernel method has been used for classification process. The highest accuracy results which obtained using the RBF kernel at sigma parameter ( ) 35, 40, 45 and 50, within 60% feature extraction are 94.79%. The highest accuracy level which obtained using the polynomial kernel within 1 parameter order and 60% to 75% feature extraction are 95.83%. The results showed that the feature extraction using polynomial kernel produces a higher accuracy than the RBF kernel method.