Brake disc positioning and defect detection method based on improved Canny operator
Jun Guo,Yang Yang,Xin Xiong,Yue Yang,Mengzhen Shao
DOI: https://doi.org/10.1049/ipr2.13026
IF: 2.3
2024-01-10
IET Image Processing
Abstract:This manuscript is based on machine vision to first position the brake disc and then detect defects. Finally, an experimental platform is built for experimental verification. The experimental results show that compared with the traditional Canny algorithm, the precision, recall, and F1‐score of the improved Canny algorithm are increased by 3.2%, 13.2%, and 6.0%, respectively. After adding parallel processing, the average detection time was reduced by 45 ms. Finally, the accuracy of scratches extracted using this algorithm reached 97.96%, the leak rate, and mistake rate of this study are only 3.33% and 2.00%, which provides a more accurate and efficient theoretical support for the research on brake disc scratch detection. Firstly, the Canny operator combined with circle fitting is used to position the working surface. Secondly, the median filter and the Laplacian operator are selected for preprocessing. Due to uneven illumination, surface texture, or environmental dust, the extracted scratches exhibit fragmentation, and traditional morphological operations cannot handle scratches well, so an improved Canny operator combined with probabilistic Hough transform is proposed for scratch connection. A bidirectional connection is introduced to optimize the edge detection process of the Canny operator, and a parallel algorithm is introduced to shorten the detection time. Finally, the standard Hough transform needs to traverse the entire parameter space, and the calculation is complicated, and the probability Hough transform is selected. The experimental results show that compared with the traditional Canny algorithm, the precision, recall, and F1‐score of the improved Canny algorithm are increased by 3.2%, 13.2%, and 6.0%, respectively. After adding parallel processing, the average detection time was reduced by 45 ms. Finally, the accuracy of scratches extracted using this algorithm reached 97.96%, the leak rate, and mistake rate of this study are only 3.33% and 2.00%, which provides a more accurate and efficient theoretical support for the research on brake disc scratch detection.
computer science, artificial intelligence,engineering, electrical & electronic,imaging science & photographic technology