Detection method for surface scratches of composite automotive components with high reflection and complicated background color

Abstract

Abstract Current testing status have the problems that the surface background color pattern of high reflective carbon fiber auto parts is complex, the shape of scratch defects is irregular, and the shallow and micro scratches are not easy to be detected. Aiming at the problems, a scratch detection method combining innovative morphological processing and optimized Canny edge detection algorithm is proposed. The image acquired by an innovative image acquisition platform. After gray processing and open operation denoising, the self-defined oval kernel secondary expansion is introduced in the morphological processing part. When threshold segmentation and minimum connected domain screening are done, the optimized Canny edge detection algorithm is used. Through the non-maximum suppression of gradient amplitude, a wide range of dual-threshold parameters are selected to detect the edge of the target image. The results show that the detection rate of the proposed method is 18.57% higher than that of the traditional way, and the detection rate is up to 95.71%. At the same time, the proposed method can reflect the scratch morphology more intuitively.