Research On Image Analysis Method Of Wood Surface Gray Defects Recognition

The existence of wood surface defects result in the decrease of wood utilization, however, the inevitable subjective factors constrain traditional manual inspection, and it is the long-term explored subject of wood industry to seek a detection and analysis method with the characteristics of intelligence, automation and non-contact. Therefore, on the basis of deeply studying on relevant literature and according to the categories and characteristic of defects, the paper proposes a method that using the image analysis technology to recognize wood surface defects. Solutions investigation, hardware system design, algorithm design and experiment research are completed.At first, through studying lighting system, the red LED strip light source is chose and the target images with high contrast and obvious defects brightness features is captured in the case of high angle illuminate system. Secondly, through analyzing the light intensity distribution curve, a gray compensation method based on exposure-reflection model is proposed to correct the original images, and then variance as the criterion for overlapped image blocking. In addition, the seven statistical parameters are calculated according to the image gray histogram, and their classification ability was evaluated comprehensively by using the distance between classes to determine the optimal statistics feature, namely smoothness feature, which can recognizes effectively the defects such as knot, decayed and polluted and so on. Finally, through analyzing the probability distribution histogram of smoothness, an adaptive clustering method with maximal between-cluster variance is presented to determine the classifying threshold value. Furthermore, the defects are recognized on the basis of threshold method.The online detection experiment of wood surface gray defects demonstrates the recognition rate is up to 99%. Experimental result shows the wood recognition algorithm, taking smoothness as wood surface gray property and adopting lustering method with maximal between-cluster variance to determine the classification threshold, not only possess theoretical reasons, also have excellent practicability and veracity. as a result, It improves the availability of wood, simultaneously lays some relative research basis for wood surface defects automatic detection technology.