A new approach for road centerlines extraction and width estimation

Abstract

Road centerlines detection and the estimation of road widths play an important role in many computer vision applications e.g. road network extraction from satellite images. Canny edge detector has many advantages over other first-order edge detection algorithm. For example, the edge of the line can make the connection degree well, the bandwidth can refine into only one pixel wide. But it can not extract the centerline. The Hough transform is a method for finding lines. In this paper, firstly, Canny transform is used to detect the edge of a line, then two parallel lines are extracted from the edges through progressive probabilistic Hough transform(PPHT). At last the parameter of the road centerline and its width is computed by the proposed new searching scheme. Intensive experiments have shown that the proposed methodology is effective in finding the road centerline and estimating the road width.