Automatic road network extraction from UAV image in mountain area

In this paper, we propose an approach to extract road network from unmanned aerial vehicle (UAV) images in mountain area. Firstly, a series images are arranged automatically based on their generation time and mosaiced to the zonal image along flight direction. Secondly, line segments are detected using a region growing algorithm. Thirdly, potential road segments are identified using parallel criteria. Each road's central axis is extracted from the region generated by the entire parallel line segments. Finally, we propose a novel algorithm to connect the broken road segments, remove short or pseudo segments with orientation- projection criteria.