A local annular contrast based real-time inspection algorithm for steel bar surface defects

Abstract A new inspection algorithm based on local annular contrast (LAC) for steel bar surface defects is presented. Because of little fluctuation of gray value inside the defects, traditional gradient detection methods could not find the defects accurately and often mistake the boundary of the bright region in center for defects. The LAC-based algorithm doesn’t have this disadvantage. Defects in images can be found due to their large contrast with local annular background (LAB). After average filter to steel bar surface image, the noise could be smoothed to a great extent but the contrast between defects and LAB changes very little. Then the LAC-based algorithm is applied to inspect defects. Experimental results show that the proposed algorithm needs only 13 ms to inspect one steel bar surface image and its detection accuracy exceeds 95%.