A Study Of Key Technology On Inspection System For Steel Strip Surface Defects

Having focused on the problem of ineffective recognition for surface defects inspecting on steel strip,the framework and software composition of the inspection system based on computer vision is analyzed.By researching the key techniques of the inspection system,the solutions such as initial inspection,segmentation and recognition are proposed.To improve the quality of digital sampling image,a method for preprocessing is presented.The method consists of histogram equalization with local enhancing and denoising by adaptive median filtering.Because the frequency of defects in steel strips is very low,the differences in Hu Moment Invariants between the defect and no defect image are used to make a method of initial inspection which can raise the efficiency of the system.After initial inspection for the defect image is selected.Then the defect edges composed by single pixel are extracted using fast kirsch edge detecting and connectivity preserved edge thinning algorithm.Finally,defect classification is recognized by extracting geometrical features like Beelinearity and Tamura texture features.The software is realized.And the experiments illustrated that the corresponding solutions are effective.