Study On Recognition Technology Of Digital Imagine Outline Based On Sub-pixel

With the high development of modern industry, automation is more widely used in ourdaily life. The product appearance is tested with the help of computer and the testingsystem based on machine vision is gradually used in daily life. In the Digital imageprocessing, the edge is the line dividing the object and background, and it’s also the base oftarget recognition and shape extraction. Before the image analysis process, edge detectionis the job that should be done. At the present stage, edge detection is one of the core topicsin machine vision area.When pursuing high product quality, people also pay attention to their portable problems.So in order to satisfy to people’s requirement, the parts size is becoming more and moresmall, testing requirements for image is becoming more strict and its higher testingaccuracy is also needed. It will cost more for testing if just improving camera resolution bychanging hardware to improve testing accuracy, because the cost increases with theaccuracy change. So software may be the convenient tool for people to solve that problem,and one possible way is to refine the former smallest imaging unit. Considering the highaccuracy in precision measurement process, edge detection of sub-pixel level is studied inthis paper.In digital imagine process, before edge extraction, the first job is to use the smoothingmethod to deal with the taken imagine. Mean smoothing and Gaussian smoothing arediscussed in this paper. Comparing with the former imagine, the edge grey value is gotafter smoothing is in larger differences, and edge-preserving is also not in a good situation.In this paper, after comparing with several classical pixel edges detection methods, Cannyis the most convenient way for using, but lacking smoothing is its disadvantage,anisotropic diffusion smoothing is used instead of Gaussian smoothing in order to make upthe disadvantages of Gaussian smoothing and to make sure the edge character is in a goodstage.After doing the job of locating edge imagine level, Sub-pixel edge detection based onmatrix, edge detection based on fitting and interpolation is introduced in this paper. Conclusion is that interpolation algorithm has the characteristics of simply operation andprogramming, and it also can reach a high accuracy level. So in this paper, the edgedetection precision can reach the5mlevel by using the method of sub-pixel subdivisionbased on interpolation.