Anchor Point based Hough Transformation for Automatic Line Detection of Engineering Drawings
Tianbo Liu, Quanbin Hua, Shuai Yuan, Lixue Yin, Gong Cheng · 2019
We propose an algorithm that connects the edge graph by anchor points in advance and then performs line detection. The algorithm can generate valid edge segments, each of which is a valid pixel chain. Unlike some traditional edge detection algorithms, which works on threshold gradient amplitude clustering to determine edge elements. Our algorithm firstly obtains the anchor points according to the edge boundary direction of the image and the gradient values, and then connects the anchor points through the intelligent routing algorithm to obtain the edge images, and generates effective edge segments, which are valid pixel chains, then, according to the half-sampled Hough transform line detection, so as to realize the detection of graphics in a straight line, The edge map generated by the previous stage always contains a clean, perfectly continuous, well-positioned single-pixel wide edge. In addition, the quality of the edge pixel chain is also very good, and then the line is fitted by the least squares method again, and the final detected line effect is significant.