Line Detection and Classification of Macroscopic Wood Images Using Hough Line Transform and Random Forest

Wood is a natural material derived from trees or other plants, characterized by a hard, tough fibrous structure. Identifying wood species is crucial in the wood industry to ensure species conformity and to meet quality and legality standards. To prevent identification errors that can result in economic losses and environmental damage, a fast and accurate wood identification tool is essential. The aim of this research is to develop an automated system for measuring the density distribution of lines and obtain the line features in macroscopic wood images. One method that has been used involves the Hough Line Transform for detecting and extracting lines, combined with Random Forest algorithm for classification. This research focuses on three wood species that have distinct line patterns in their appearance. A confusion matrix was used to represent the performance of the model, with 70% of the data used for training and 30% for testing. All three wood species showed a very high accuracy when using this classification model. Lengkeng had the lowest accuracy at 93%, while Meranti and Puspa had the highest accuracy at 97%. The evaluation showed that the model accurately classified the line density in the macroscopic wood image, yielding final classes such as very close, close, wide, and very wide.