Study On The Segmentation Of Zipper Teeth From The Zipper Product Image

The visual inspection of the zipper is a key steps of the production.The traditional detected method of the zipper is implemented by man,and is heavy-worked,inefficient and low-accurate.For this reason,it is necessary to study how to detect in zipper products by image detection of defects in order to achieve greater efficiency and accuracy,reduce labor intensity of workers.Teeth-missing is one of the main type of defects in the visual inspection of the zipper defects.To judge teeth-missing drawback by the image detection,we need to separate the area of zipper teeth from the image.This paper on the basis of analyzing the appearance of the zipper images,studies of the use of GLCM and RKM for the process of the segmentation of zipper teeth deeply.Firstly,this paper analyzes the status of the detection of zipper image,the status of the analysis of texture image and the status of the segmentation of rough set.Secondly, this paper adopts the preprocesser of the zipper image,including image grayscale,image enhancement,for completing compensation of the light of image, smooth and so on,with the ultimate aim to provide the image data as effectively as possible for the next treatment.Thirdly,this paper based on the basic of the theory about the extraction of texture feature designs the algorithm of using GLCM to extract the texture feature of zipper. On this basis,this paper builds a feature matrix,using the RKM algorithm to cluster the feature matrix,making the pixels in the area of zipper teeth cluster into a single category.Fourthly,this paper uses the algorithm designed to process images on the zipper, analysising zipper texture feature data extracted by GLCM, discussing the adaptive algorithm of segmentation of zipper teeth from different zipper product,studying of the neighborhood size,the number of random points on the affection of the segmentation of zipper teeth.The experiments show that the texture descriptional method and extractional method can rapidly and accurately extract the zipper teeth from the image,and need not to modify to extract the zipper automatically when the product images change in the zipper.The number of random points and neighborhood size does not affect the final extraction of the zipper teeth in the RKM algorithm,which shows the method is adaptive and robust.Finally,this paper summarizes the work conducted and gives suggestion for further research.