Incorporation of pre-classifier and nonlinear feature selection for tropical wood species recognition system

Automatic classification of tropical wood species is becoming more important especially for timber exporting countries due to the considerable economic challenge as a result of fraudulent labelling of timber species at the custom checkpoints. Hence, a reliable automated wood species recognition system is needed to inspect the wood species labelling at the checkpoints. A tropical wood species classification system is designed based on image analysis. There are thousands of images being processed in the wood database. In this paper, the incorporation of the nonlinear feature selection with fuzzy-based pre-classifier is proposed in order to simultaneously solve the nonlinear problems and solve problems due to large wood database which will then improve the performance of the wood species recognition system. Besides that, the proposed pre-classifier which emulates the experts' way of inspection enables human intervention when monitoring the wood species stored in the wood database. The research involves comparative analysis of the system for different configurations of fuzzy-based pre-classifier. The results show that the proposed configuration of fuzzy pre-classifier and nonlinear feature selection gives better result compared to other configurations in terms of classification accuracy when more wood database is added into the system.