Extraction of Urban Roads Based on Edge Information from High-resolution Satellite Imagery

摘要

This paper proposes the method of urban road extraction from a high spatial resolution satellite image. The proposed method is composed of three steps: detection of the dominant road directions, extraction of the initial road lines, and extraction of the final roads. In the first step, the dominant road directions are extracted by accumulating the length of each Hough line according to its direction. With the dominant road directions and edge information of the multi-spectral images, the initial road lines are found. The final roads are extracted from the initial roads by filtering. The proposed algorithm evaluated its efficiency by applying the Quickbird-2 image. The completeness and correctness of the result from the proposed method were 68% and 62%, respectively. This study shows the efficiency of the edge information of multi-spectral images to detect the urban road network using only one satellite image.