Font Size: a A A

Research And Application On Feature-based Image Defect Detection Method

Posted on:2012-03-29Degree:MasterType:Thesis
Country:ChinaCandidate:X G YeFull Text:PDF
GTID:2218330362457481Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the progress of manufacturing and industrial production processes,the quality of product is increasingly demanded. The defection detecting system which based on machine vision become an important measures of quality control.With the rapid development of machine vision,the detecting precision and speed become a bottleneck of machine vision development, In order to overcome this difficulty, present an image defect detect algorithm which based on image feature.In order to improve the stability of the feature detection algorithm, present two kind of improved pretreatment algorithm.To solve conventional image enhancement method often lack of adaptability, bring forward an image enhancement algorithm which is dynamic and fast.According to gray scale distribution, stretch gray scale to the ideal position automatically,the biggest image gray-scale pixel of original image stretching to 255,and the The minimum pixels stretching to 255. other gray-scale gray distribution stretching accord to the original proportion, choose four typical image for algorithm experiment ,then process image enhancement to low and brightnessimage respectively. The experiment results show that the enhancement effect for differend brightness and types image is ideal ,compared with the conventional enhancement mathod,have good adaptability.Aim at the shortcomings of image blur caused by conventional image denoisiong method, Smart filtering algorithm proposed,according to whether the D-value of one pixel and its neighbor pixel between at a threshold range,decide if filter the pixel. This algorithm can remove noise pixels, meanwhile retaining image edges, The comparsion filtering experiment between common median gaussan filtering and smart filter methods,by chooosing an image which have rich edge information. Experiment results shows,compared with the conventional filter method,smart filter edge-keep ability quite ideal,and has lower Time complexity.According to conventional image detection methods cannot distinguish the image defect type,proposed a detect method which based on feature.Choose and extract the detect feature set from test image according to its defect classfication,Analysis and compare test feature one by one,to determine whether the defect feature set were completely matched, fetch test result.As less feature data were compared, the purpose of quick and accurate detection methods was realized. Algorithm experimen results show that feature detection algorithm can classfy image defect by compare the image feature information. the algorithm hava high efficiency in solving problem between accuracy and speed of image detection, has higher value of using.
Keywords/Search Tags:Image Feature, Defect Detect Printed image, Machine vision
PDF Full Text Request
Related items