Abstract: In order to meet the requirements of deep subpixel localization and strong noise resistance for image edge detection in machine vision precise measurement,a Canny-Devernay subpixel image edge detection algorithm based on bicubic interpolation was proposed. Firstly,the median filtering algorithm was used to reduce the noise in the acquired images. Secondly,the image edges were subdivided through bicubic interpolation. Finally,the contour generation step was optimized and Ostu algorithm was applied to obtain an adaptive threshold,which was passed into Canny-Devernay algorithm to achieve accurately extracting subpixel edges of image. An experimental system was developed by high-resolution industrial camera and high-performance computer,and experiments were carried out on the OpenCV image and the acquired images of USB interface plug-in parts as samples,which was done with camera calibration accuracy of 0.009 8 mm/pixels.The experimental results show that the average edge detection error of the proposed algorithm is 0.006 85 mm,which is less than 0.7 pixels,and the calculation time deviation is 7.68 ‰. The edge positioning accuracy,noise resistance and compute stability of the proposed algorithm were better than the Canny algorithm,the algorithm based on Zernike moments and the Canny-Devernay algorithm. The proposed algorithm can better meet the requirements of the stability,reliability and high-precision image edge detection for machine vision precise measurement,and would be used to develop a new type of machine vision dimensional precise measurement device.

Key words: machine vision, edge detection, bicubic interpolation, binary threshold with hysteresis, Canny-Devernay algorithm, subpixel