Research Of Surface Defects Detection Algorithm Based On Machine Vision

Abstract:The significance of quality control in production is more and more notable. The surface defects of product affect its quality. So, the detecting of product surface defects plays an important role in improving its quality. The traditional manual detecting method can not satisfy the demand due to the labor-intensive and low detection efficiency. Therefore, automatic detecting methods of surface defects be studied.This paper researches the detecting technology of surface defects based on machine vision. It focuses on the image filtering and denoising algorithm, and discusses methods of segmentation and determination of defects. The main work be summarized as follows:Regarding a large amount of noise in images which cause a great deal of interference in defects detection, the paper studies the effect of several commonly used filtering algorithm for image denoising, and analyzes its shortcomings, then proposes the sequential filtering denoising method. This method not only can remove the noise but also can enhance the target information to be detected, thus it lays a good foundation for the segmentation of the defects. For linear defects, it compares results of commonly used real-time edge detecting method and line detecting method. On this basis, it improves the line detection method. This paper studies intermittent linear defects which are common but difficult to detect.It designs a method of connection of discontinuous segments based on Probabilistic Hough Line Detection. Experiments show that this method can effectively connect intermittent segments. For stripe defects, it analyzes the shortcomings of the classic detection methods and puts forward Peak-valley-Difference based on projection. Experiments show that this method can effectively detect stripe defects. For collapse edge defects, it raises a detecting method based on projection. Experiments show that this method is effective.Finally, it summarizes the main research work, and brings forward the idea of the follow-up study tasks. There is guiding significance for future research.