Keywords: 
• forest engineering /
• gray level co-occurrence matrix (GLCM) /
• gabor filter /
• invariant moment /
• wood image /
• defect detection

Abstract: It is important to detect the wood surface defects using computer vision technology. In this paper，a defect detection method which can accurately and robustly determine whether there is defect on wood surface image or not is proposed based on mixed texture features. At first，gray level co-occurrence matrix(GLCM)，Gabor filtering and invariant moment method are used to extract 10 image scale，translation，rotation invariant and texture features optimally. Then，feature vectors are mixed effectively. Finally，BP artificial neural network is used to train the sample sets and detection based on the mixed texture features. Experiments show that the proposed method can detect surface defects of wood boards accurately and the average success rate of detection is 96.2%.［Ch，4 fig. 1 tab. 12 ref.］