Development of defect detection algorithm for surface inspection system using twin illumination and subtraction technique

Abstract

Surface inspection of steel products is very important for quality assurance. Many automatic in-line surface inspection systems using camera techniques have already been installed in steel sheet production lines. However, automatic surface inspection of steel products such as steel pipes and plates has not advanced because the entire product surface is covered with uneven mill scale, and it is difficult to distinguish the pattern of the mill scale from defects with concave-convex shapes in images captured by a camera.

The authors developed a new surface inspection system using the twin illumination and subtraction technique, which emphasizes only concave-convex defects while canceling the pattern of the mill scale on steel pipe and plate products. An optical approach to enhance the detection performance of this system was already reported in connection with the development of a steel plate surface inspection technology. This paper discusses the development of a new image processing technique which makes it possible to detect defects from obtained images that satisfy the conditions for introduction of automatic inspection systems at steel production lines. Concretely, a high-speed bright-dark pattern detection algorithm was developed by using expansion and conjunction processing, and the processing speed was improved by 18.4 times relative to the conventional simple method. An effective new original feature, overlap ratio of bright parts, was also developed, resulting in a 5.15 % improvement in the classifier concordance rate relative to that without the proposed feature. These developments have realized automatic surface inspection systems suitable for introduction in steel product manufacturing processes.