An edge-detection method based on adaptive canny algorithm and iterative segmentation threshold

• Published in: 2016 2nd International Conference on Control Science and Systems Engineering (ICCSSE) pp. 64 - 67
• Main Authors: Song Qiang, Lin Guoying, Ma Jingqi, Zhang Hongmei
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2016
• Subjects: Algorithm design and analysis; Canny algorithm; crack detection; edge detection; Filtering algorithms; Filtering theory; Image edge detection; Image segmentation; iterative method; Surface cracks; Surface morphology
• DOI: 10.1109/CCSSE.2016.7784354

In the technology of crack detection, the traditional Canny algorithm uses fixed spatial scale coefficient of Gauss filter and empirical values of the high and low thresholds, and it has defective in self-adaptability because it is unable to correct parameters according to the actual image. This paper proposes a method based on adaptive Canny algorithm and iterative threshold segmentation algorithm to detect surface crack. Firstly, in the process of image smoothing, the adaptive Canny algorithm automatically calculates difference between the gray value of present pixel and the average gray value of image in the filter window, and the difference is set as the spatial scale coefficient for Gauss filter of the current pixel. Secondly, the Otsu method is applied to calculate image gradient histogram so as to get high and low thresholds and those new values are applied to detect the edge of the crack. Finally, based on the iterative threshold segmentation algorithm, the image with crack is binarized, and the breakpoints in the crack and the cavities inside the crack are eliminated by morphological dilation. Experimental results show that the new crack detection method can well retain crack edge and get better effect on noise cancellation. It can also reduce the false and missed detections. Better results can be achieved on crack detection.