Sub-pixel Level Edge Extraction Technology for Industrial Parts for Smart Manufacturing

Abstract

With the development of science and technology, image edge detection and extraction technology, which is one of the most basic and important aspects of digital image processing, is gradually being applied to life science. Currently, image edge detection and extraction for sub-pixel lacks accuracy, which suffers from image noise such as shadow burrs or distortions. To address such challenges, in this paper, we use histogram equalization, bilinear interpolation, and progressive and least-squares fitting methods to extract the high-precision edge contour at a sub-pixel level from complex images with interference noise and aberrations in the edge contour and analyze it. Thus, we obtain a new image with uniform grayscale distribution, and no shadow interference or distortion, to extract subpixel images with higher accuracy compared with the original image using bilinear interpolation. And then we extract subpixel edge contours and obtain contour data (number of contour subpixels, contour length) by the improved Canny operator and the FindContours function of the OpenCV library. In this way, the edge information of parts can be extracted from the original fuzzy boundary in smart manufacturing.