Strip Surface Defect Edge Detection And Segmentation

In the automatic defect detection of the surface of the metallurgy strip, the computer vision technology plays an more and morn importance role. During the detection process, one of key factors is image segmentation in which the most important step is detecting the edge of defection to obtain the defection features. Previous studies have shown that the great difficulty confronted in the image processing measures and theory, in the low and middle-level, become the fundamental matters, which determine how well the matter was understanded by the system of the computer vision online automatic defect detection, and a lot of problems need more research to solve. Low-level image processing maily include image noise reduction, contrast enhance,sharpping,restoration , while the middle-level image pracesing include the edge detection,segmentation,description of object feature and so on.Generily spaking, there are two schemes of image edge detection: one is based on the feature of the image grey level, the other one based on the multi-scale wavelet. The former classical edge detection operaters have been developed, but the applicability and the ability of resisting noise is limited ,while the later abstract the edge information by multi-scale wavwlet transform and the Lipschits index of the singal and noise under the wavwlet transform which reducing the disturb of the noise.In the first chapter, the technique about the detecting strip surface is introduced and the feasibility of wavelet applied in this area is analyzed. Chapter 2 analyzes characters of this type image and the model of the defect edge using the imaging process of the metal strip suface. After introducing the classical edge detection operaters, the corresponding experimental results are exhibited.Chapter 3 introduces the wavelet basic theories and the application of the means of removing noise based on the wavelet shrink and the phase consistency restriction in the system of automatic detection for mental strip suface. Chapter 4 discusses the wavelet module maximum algorithm and multi-scal adaptive threshold algorithm for detecting the defect edge. The detection results show that the algorithms work well in abstracting defect edge on the mental strip suface and accord with the charatictistic of human vision sestem. Chapter 5 presents the defect detectional mesures based on the statistic which also has a good result on the defect detection of the the mental strip suface. .May all the work in this paper is of some value to research and application of the techniques for detecting defects on the surface of the mental strip.