Wood Species Identification Based on Gray Level Co-Occurrence Matrix (GLCM) Features on Macroscopic Images

Wood is an incredibly valuable resource, particularly for everyday living. To fully harness the advantages of wood, it must focus on two key considerations. Firstly, it is imperative to consistently utilize wood sourced from sustainably managed forests. Secondly, we must explore techniques that maximize the utilization of every part of the tree. One technique for meeting these considerations is to create a wood identification system. This system can be used for quickly inspecting wood species. In wood identification, it is essential to consider specific characteristics and physical properties of wood. Manual identification will depend on the examination of wood anatomists’ eye and will require a significant amount of time. In accordance with these situations, a computer vision-based system can address this condition. Therefore, feature extraction is necessary to extract the features of wood characteristics from the wood image. This research aims to propose a method for wood species identification based on Gray Level Co-occurrence Matrix (GLCM) features to extract important information about wood characteristics from macroscopic wood images. For the classifier, the Random Forest algorithm is proposed for the identification of the machine learning model. Five wood species images will be used in this research, with each wood sample being presented as a macroscopic image. The total dataset used was 750 images, with each wood species having 150 images. The result showed that the Model C (90/10) training data ratio demonstrates good performance in classifying wood species from the macroscopic images. The model achieved a peak accuracy of 0.81 and correctly predicted all test images. This study indicates that the Random Forest model can be an effective classifier for wood species identification.