A texture surface defect detection method and system based on anomaly synthesis and decomposition

Abstract

The invention discloses a texture surface defect detection method and system based on abnormal synthesis and decomposition, and belongs to the field of image processing. The invention constructs a segmentation-guided defect generation network, and can use a small number of real defect training samples to generate a large number of defect samples similar to real defects. At the same time, an abnormal synthesis method based on Gaussian sampling is proposed, which can use defect-free positive samples to randomly synthesize abnormal negative samples, which can solve the problem. The problem of a small number of defect samples in the industry is solved, and the defect detection accuracy is further improved; the invention uses an abnormal decomposition network to decompose abnormal negative samples into texture background images and abnormal mask images, which can effectively prevent defects from being reconstructed into the texture background. Improve the texture background reconstruction accuracy, and accurately segment the defect area, and fuse the residual image with the abnormal segmentation image, thereby improving the defect detection rate and reducing the defect over-inspection rate.

Description

technical field [0001] The invention belongs to the field of image processing, and more particularly, relates to a texture surface defect detection method and system based on abnormal synthesis and decomposition. Background technique [0002] In the field of industrial manufacturing, the quality of raw materials of various industrial products varies, and the manufacturing process is complex. Various surface defects will inevitably occur on the surface of products, such as textiles, new display devices, ceramics, steel, etc. Surface defects refer to localized areas that differ from the surrounding texture and pattern, or localized areas with irregular brightness variations. Since the surfaces of various products often show different texture characteristics, these texture surface defects will directly reduce product quality and affect user experience. To improve production quality, all types of surface defects should be strictly controlled during the manufacturing process. ...

AI Technical Summary

Problems solved by technology

Existing methods can only be adapted to one or several types of textures (for example, they can only detect display devices, but cannot be applied to wood surfaces), and can only detect fixed types of texture defects (for example, they can only detect high-contrast defects, but cannot detect low-contrast defects. defects), it is difficult to solve all cases

Images