Research On Wood Surface Defect Image Segmentation Based On Computer Vision

It is a very important topic of the detection and classification of wood that to detect and classify the surface defect of w ooden plate. The technol ogy of detection of Wood-line surface defect based on m achine vision cost less, and less in fluenced by the environment. It is very adaptable to diversity of wood, and has a high level of robustness and accuracy. This kind of technology is suitable for the flexible m anufacturing systems of wood. Because of these advantages, it has more and more research on how to apply the machine vision in the detection of surface defects of wood in recent years.This paper explores the algorithm of image processing and image segmentation about three types of surface defects of wood, such as wormhole, live knot and dead knot. Main works and contents are as follows:Firstly, we determine whether it is exist defects in collected wood image. We dissect the curve of histogram of wood i mage, observing whether there is an exhibit change of wave. If there are two peaks, and large differences in height of the primary and secondary wave, we determine there are surface defects in wood image. If there is only one peak in the histogram, it might be normal images of wood. After make it is true that there is exist the surface defects of wood, acco rding to ch aracteristics of the im age of wood, we sharpening enhancing the image, and undoing noise of the image. We used median filter to wiping off the yawp for im age. Because de-noising processing will blur the border of image, we need to enhance the boundary information with gradient method.Secondly, we will segm ent the pre-proc essed image of wood. W e use the Ots u method to segment the image based on genetic algorithm, this kind of algorithm build a system model to f ind the optimal solution by simulates biological evolution processes. Because there are many independent interference information in segmented image, these affect the extraction of im age edge, we need to use the m ethod of m athematical morphology to reprocess the segm ented image. We consider using open operation to remove small connecting in segm ented image by dilation and erosi on, such as isolated point, burr and protruding parts of the image, cut the connection of slender to segment the interest regions of i mages. We compare th e characteristics between the best g lobal thresholding algorithms on Otsu m ethod, region growing algorithm and Otsu m ethod based on genetic algorithm by experiment.Additional, we use the Canny operator to extract the defect edge of segmented wood image. Canny operators have a good position ing performance, it will n ot miss the real edge.We verify the feasibility of the algorithm by experiment. Experimental results show that this m ethod makes the wood defect c ontour smooth and clear, coherent by using genetic algorithm and mathematical morphology algorithm to segment the image of wood defect. This algorithm provides the basis for recognition.