Design of an Intelligent Wood Species Recognition System

Tropical rainforest has more than 3,000 different types of timber species. According to the Forest Research Institute of Malaysia, out of these about 200 species are being used by the timber industry. Among the major timber consumers are housing developers, wood fabricators and furniture manufacturers where the need for recognition of wood species is necessary. Automatic wood recognition has not yet been well established mainly due to lack of research in this area and the difficulty in obtaining the wood database. In this paper, an automatic wood recognition system based on image processing, feature extraction and artificial neural networks was designed. The proto-type PC-based wood recognition system is capable of classifying 30 different tropical Malaysian woods according to their species based on the macroscopic wood anatomy. Image processing is carried out using our newly developed in-house image processing library referred to as “Visual System Development Platform”. The textural wood features are extracted using a co-occurrence matrix approach, known as grey-level co-occurrence matrix. A multi-layered neural network based on the popular back-propagation algorithm is trained to learn the wood samples for the classification purposes. The system can provide wood identification within seconds, eliminating the need for laborious human recognition. The results obtained show a high rate of recognition accuracy proving that the techniques used is suitable to be implemented for commercial purposes.