Line Detection for Point Set of Varying Discrete Degrees

Abstract

Many line detection algorithms based on Hough transform are used to work on continuous set of points, such as edge image, which is composed of continuous or almost continuous points. When these algorithms work on the high-discrete point set, they produce the multi-short-line segments because of the noise and large gaps between points, which seldom characterize the whole line segment completely. We proposed an improved HoughLines algorithm to solve this problem: First, we propose a new sorting strategy of the point coordinates that is self-adaptive for each line segment; second, we improve the strategy of finding the non-background points that contributed to peaks of Hough transform matrix through a tolerance threshold. The improved algorithm can detect line segments in point set of varying discrete degrees at a high precision. A number of experiments show our method is very efficient.