Vision-based surface defect inspection for sequence circular objects

High-speed defect detection method must be required for on-line inspection as well as accuracy based on vision technology. An appropriate real-time defect detection and location algorithm was investigated on surfaces of sequence circular objects in order to be used for on-line detection system. Firstly, difference image method and Otsu segmentation were employed in extracting defect regions. Secondly, two-steps location (TSL) method was presented for objects to find which object defects belong to. Specifically, gray projection was served as a skeleton location for circular objects; what´s more, improved canny algorithm based on segmented arcs combined with least square method (LSM) is used for accurate location. Experimental results indicate that the algorithm researched in this paper is faster than Hough transformation and morphology algorithm, which only needs 0.22 seconds for defect detection. Additionally, it can not only effectively detect defects on surfaces of objects but also calculate their positions. Therefore, it can be appropriately used for on-line defect detection.