Research On Image Segmentation Algorithm Of Wood Surface Defects

Wood surface defects would seriously affect its quality, performance and use value, so the segmentation and detection of wood surface defects has important meaning for lumber grading evaluation and quality control, raising the utilization ratio of timber, saving existing resources. At the same time,it is an important step in the process of plate production. This paper makes segmentation experiments of wood image whose surface contains three kinds of typical defects,such as bugs, slipknots,and promulgates. The main methods are as follows:The traditional segmentation algorithm is studied, including the theory of edge detection, threshold segmentation, image segmentation based on region.We select many images which contain a single target of bug, slipknot or promulgate, and use the traditional segmentation algorithm to make wood surface defects segmentation experiments and result analysis.We make theoretical research and experiments of traditional C-V model, traditional SNAKE model, GVF SNAKE model, traditional GAC model. Aiming at the shortcomings of the traditional model segmentation algorithm, we use C-V model combined with morphology, improved algorithm of C-V model, improved algorithm based on the combination of wiener filtering and GVF SNAKE model, the improved algorithm of GAC model to further study images segmentation of wood surface defects, including the following parts:(1) Firstly we make preliminary segmentation experiments of wood images which come from picture gallery first and contain a single target, and find out the influencing factors of segmentation results, we also make experiments of improved algorithms and verify their feasibility; (2) Secondly we make the comparative experiments and analysis of the same group of wood images with a single target defect, and verify the superiority of improved algorithms; (3) Finally, we use improved algorithms to make segmentation experiments of the wood images which come from picture gallery second and contain multiple defect targets, and verify the practicability of improved algorithms.We Select images which come from picture gallery second and use various improved algorithms to make contrast experiments and result analysis, then compare the segmentation time required, the complexity of the segmentation process, the final rendering and the anti-jamming performance of noise and so on, and choose a more superior wood surface defects segmentation algorithm.Use the selected algorithm to make multi-objective defects segmentati-on experiments of the whole piece of wood images in complex background which come from picture gallery third,and also add coordinates to effectively find the location of wood defects.