Image Feature Extraction Method of Strip Surface Defects Based on Local Feature Space Distance

Abstract

The invention relates to a strip steel surface defect image feature extraction method based on local feature space distance. The technical solution is: for a preprocessed vector data point Xi of a strip surface defect image, select from the matrix data X formed from all the preprocessed vector data points of the strip surface defect image and a strip surface defect After image preprocessing, the K neighbor points of the same category of vector data points Xi establish the manifold local feature space S(Xi), and then measure the multi-manifold divergence Js by the distance between the manifold local feature spaces S(Xi), On the basis of keeping the local structure of the manifold unchanged, the multi-manifold divergence is maximized to find the low-dimensional projection matrix A, and the discriminant feature extraction of the strip surface defect image is realized. The invention extracts the classification features of the surface defect image of the strip steel by maximizing the divergence Js of the multi-manifold, and has the characteristics of improving the recognition effect of the surface defect image of the strip steel.

Description

technical field [0001] The invention belongs to the technical field of strip steel surface defect image feature extraction. In particular, it relates to a strip steel surface defect image feature extraction method based on local feature space distance. Background technique [0002] Strip steel is one of the main product forms of the iron and steel industry. It is an essential raw material for aerospace, automobile and ship manufacturing, etc., and is related to the development of many manufacturing industries. In recent years, the demand for strip steel has been increasing and requires a high surface quality. In the rolling process, due to continuous casting billet, rolling equipment and rolling process, etc., defects such as cracks, scale, scabs, roll marks, scratches, holes and pitting appear on the surface of the rolled steel plate. , These defects not only affect the appearance of the product, but more importantly, reduce the performance of the product such as corrosio...

AI Technical Summary

Problems solved by technology

However, due to the wide variety and complex features of strip surface defects, the recognition rate based on these traditional feature extraction methods is not ideal.

Images