Urban road extraction on the dsm data based on ART, hough transform, and B-Spline

In this work, we describe an approach to road extraction in an urban area. In urban areas, the road is dominant and surrounds the buildings. In aerial photos, building shadows are the main problem for road extraction. In this case, we use the Digital Surface Model, which is based on the elevation of land surface, to minimize this problem. We propose the threshold selection based on ART clustering in the geometrical histogram point and combined with the Hough transform as the cluster evaluation. This thresholding is for extracting the candidate road. The region growing is then carried out to expand the road, followed by a morphology operation to fill in the holes and remove large areas. Finally, the road lines are constructed based on the B-spline curve. The experimental results show that the proposed method improves the quality with good performance.