Application of an improved fast algorithm in line detection of parts

Abstract

In order to solve the problem of Hough Transform for line detection, an improved fast algorithm for the detection of multiple lines had been studied out. Based on the idea of "partially Hough Transform goes first, and then wholly least squares fitting", the influence of quantization error in the digital image has been reduced, so that the straight line can be positioned accurately. Resetting the neighborhood of the detected line by using the minimum distance criterion can solve the problem of the uncertainty of random Hough Transform effectively, as well as the problem that the remaining point set will interfere with the subsequent line detection for multiple lines detection. Some main issues in the algorithm had been analyzed; the relation curve of threshold selection and errors in the detection results has been made. Dealing with some linear parts by this, it shows that the algorithm has high accuracy, timeliness and flexibility.