On study of a method for detecting micro-deformation defects of steel plate surface

In order to solve the problems such as the high difficulty and low detectable rate of the micro-deformation defects ascribed to steel plate surface deformation, this paper proposes a method of deformation defect analysis via texture collecting. First, carry out image analysis on the working principle of different camera lighting methods on the steel plate surface, and construct an image acquisition system based on the characteristics of bright and dark ground illuminations. According to the diffraction of light occurring on the steel plate surface, use industrial camera to collect the surface texture that is similar to raster to analyze micro-deformation defects. Secondly, carry out grey level transformation against the images collected via histogram equalization method to eliminate the influence of uneven illuminations, smoothen the boundary with morphological processing and eliminate the gray scale change in the striped area via Otsu binarization. Finally, propose three characteristics for the pre-processed image for characteristic extraction according to the gradient information of the image on the pixels and use the improved SVM as a classifier and also the extracted features for the purpose of image dichotomy. We also studied the influence of the eigenvalues on the accuracy of defect detection under different sizes and effective values. According to the experimental results, this method is able to detect micro-deformation defects of steel plate surface quickly and accurately, without relying on sensor to measure depth information of the surface.