Feature extraction of workpiece circular contour based on Sobel operator

In the process of circular radius measurement of binocular vision workpiece, the circular contour feature extraction is the key of space arc reconstruction. Due to the texture of the workpiece surface, the interference of the surrounding environment, uneven illumination, the existing algorithm can not accurately extract the circular contour feature. In this paper, based on the Sobel operator edge detection, firstly, the fusion gray value is generated by adaptive convolution operation and DLBP texture feature, and the edge contour points are further screened by using the fusion gray value to the edge points detected by Sobel operator. At the same time, the polar coordinate distribution histogram is obtained and the normal distribution is carried out. It eliminates the interference of noise at complex background and can distinguish the outer contour feature and inner contour feature. The experimental results show that the proposed algorithm not only overcomes the influence of illumination, but also has good accuracy and robustness.