Strip steel surface defect classification method

Abstract

The invention discloses a strip steel surface defect classification method. The method is carried out on the basis of a Gabor feature fusion block histogram; firstly, a sample library is established,and a sample set is divided into a training set and a test set; secondly, all pictures in the sample set are normalized to be same in pixel size and are imported into a Gabor filter for performing multi-scale multi-direction filtering to obtain filtering result graphs; thirdly, the filtering result graphs are fused into result graphs with the number as same as a scale quantity, the result graphs are divided into blocks, dimensionality reduction is performed on each sample picture by using KPCA, and an SVM classification prediction model is built by using labels of the sample pictures in the training set and the corresponding feature dimensions; and finally, a PSO algorithm is used for performing parameter optimization on an SVM, classified prediction is performed on the SVM model by utilizing the processed sample pictures in the test set, and then to-be-tested strip steel pictures can be subjected to comparative test. The method is high in automatic classification accuracy and high inexecution efficiency, and has huge economic values.

Description

technical field [0001] The invention belongs to the field of visual recognition, and relates to a method for classifying and recognizing surface defects by using visual recognition, in particular to a method for classifying surface defects of strip steel. Background technique [0002] Nowadays, strip steel products are used more and more widely, and the quality requirements for strip steel surfaces are getting higher and higher in industries such as automobiles, machinery, furniture, and aerospace. The quality inspection of strip appearance is an important part of strip quality. It affects the appearance of strip steel and the grade evaluation of product quality, which can directly affect the core competitiveness of products. However, at present, the detection method of strip steel defects in my country has been using manual uncoiling sampling inspection or stroboscopic light detection for surface quality inspection. Probability calculation and human judgment to determine. ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a method for automatically classifying the surface defects of strip steel to solve the problems of low classification accuracy and low efficiency in the prior art. The present invention classifies the surface defects of strip steel through machine vision , in order to achieve the function of defect identification and classification, meet the identification requirements, specific economic value

Images