Systematic feature analysis on timber defect images

Abstract

Feature extraction is unquestionably an important process in a pattern recognition system. A clearly defined set of features makes the identification task more effective. This paper addresses the extraction and analysis of features based on statistical texture to characterize images of timber defects. A series of procedures including feature extraction and feature analysis was executed in order to construct an appropriate feature set that could significantly distinguish amongst defects and clear wood classes. The feature set is aimed for later use in a timber defect detection system. To assess the discrimination capability of the features extracted, visual exploratory analysis and statistical confirmatory analysis were performed on defect and clear wood images of Meranti (Shorea spp.) timber species. Findings from the analysis demonstrated that utilizing the proposed set of texture features resulted in significant distinction between defect classes and clear wood.