Research On Image Segmentation Algorithm Of Wood Surface Defects

Wood surface defect image segmentation is an important link in wood processing,which can effectively improve the utilization and economic value of wood,and save existing forest resources.In this paper,three kinds of wood surface defect images as the research object,including slipknot,wormhole,encased knot,research on single and multiple targets and the whole plate surface defect image segmentation algorithm.The research contents are as follows:Segmentation experiments have been implemented to typical defect images by three kinds of the traditional detection method,including threshold method,region method and edge detection operator.Segmentation result is analyzed and and find out the influencing factors of segmentation results.The traditional LBF model(Local Binary Fitting)algorithm and its improved algorithm were used to compare the parameters of the segmentation results.The feasibility and superiority of the improved LBF model algorithm are verified by using the segmentation time,the complexity of the segmentation process and the integrity of the two models.The traditional Graph Cuts algorithm and GrabCut algorithm(improved Graph Cuts algorithm)are used in the theoretical research and segmentation experiments.Aiming at the deficiency of GrabCut algorithm,the GrabCut algorithm is improved.Firstly,the SLICO algorithm is used to segment the image of the wood surface defect,and then the image is segmented by the GrabCut algorithm.The superiority of the improved GrabCut algorithm is verified according to the time of the GrabCut algorithm,the complexity of the segmentation process and the integrity of the segmentation algorithm.The superiority of the improved GrabCut algorithm is verified by the segmentation time,the complexity of segmentation and the integrity of the two algorithms.The advantages and disadvantages of the five freguently used method,such as threshold method,edge detection operator method,region segmentation method,energy function segmentation method and graph cut method,are summarized,and confirm the best method.On the basis of the above research,the above segmentation methods are integrated into an experimental software system.