Road extraction from high-resolution remotely sensed image in dual space

Road extraction is one of important research contents in remote sensing field. One kind of road-detecting algorithm in dual space is proposed aiming at high-resolution remotely sensed image with complex background. This algorithm includes the following steps: Firstly, mapping both the pixels gray standard deviation and the grade vector mean of the line objects separately into the dual space by project transform. Secondly, detecting road objects based on the distribution rules of the peak and the valley formed by the two line features of standard deviation and grade mean in dual space. Thirdly, searching and tracking the road objects around the detected locations, and extracting the road network. This proposed algorithm is insensitive to the contrast between road and background, and can detect the roads from the RS images with complex background. And the experiments also indicate that the proposed algorithm is efficient for extracting road network from remotely sensed image.