The Research Of Strip Steel Surface Defect Detection Methods Based On Structure-texture Decomposition Model

As an important raw material in aerospace,machinery,automobile and other industries,the quality control of strip steel is particularly important.In the process of processing,casting and rolling,there are many kinds of defects on the strip steel surface,which not only affect the appearance of strip steel,but also reduce the corrosion resistance and fatigue strength of related products.With the development of industrial automation and intelligent technology,surface quality detection technology based on computer vision can effectively improve product quality and production efficiency.Aiming at the problems of miscellaneous defects and low contrast between defects and background of strip steel surface,two methods for defect detection of strip steel surface are proposed in this paper,including the defect detection method based on contrast pre-adjustment and image decomposition,and the defect detection method based on self-reference template guided image decomposition.The main contents of this paper are as follows:(1)Traditional texture analysis methods are susceptible to noise,and are ineffective for the types of defects that are similar to the background texture and have low contrast.In order to extract the overall structural features above the texture level,a defect detection algorithm based on contrast pre-adjustment and image decomposition is proposed.Firstly,the two contrast pre-adjustment methods proposed in this paper can compress the dynamic range of background brightness of strip steel image and enhance the contrast between defect and background,which is helpful for subsequent image decomposition to extract the overall structural features of defects in strip steel surface image.Then,the two kinds of enhanced images are decomposed into structural component and textural component by using the image decomposition model based on total variation,in which the structural component contains complete defects and the defects differ greatly from the background.Finally,the defect region can be located in the structural component by adaptive threshold.The experimental results show that the proposed algorithm can accurately detect various types of strip steel surface defects and achieve pixel-level detection accuracy.(2)In order to introduce the characteristics of a large number of defect-free images into the process of image decomposition,this paper presents an algorithm of strip steel surface defect detection based on self-reference template guided image decomposition.Firstly,according to the statistical characteristics of a large number of defect-free strip steel images,the background brightness of each image is modeled according to its brightness distribution,and a specific self-reference template is generated for each test image.Then,with the guidance of self-reference template,the test image is decomposed into structural component and textural component by using the image decomposition algorithm based on total variation.Here,by introducing a new index,gradient similarity,to measure the similarity between the self-reference template and the textural component,the parameter of image decomposition is optimized.Finally,the decomposed structural component is enhanced in frequency domain and the defects are located by adaptive threshold method.The experimental results show that the proposed algorithm can be applied to various types of strip steel surface defects,including tiny defects and low contrast defects,and obtain higher detection precision,recall and F-measure.At the same time,the algorithm can also be extended to other uniform or periodic texture surface defect detection fields.