Road Network Extraction From Low-Contrast SAR Images

In low-contrast synthetic aperture radar (SAR) images, the contrast between roads and surrounding objects is low; therefore, many false roads will be introduced in the process of extracting road networks. To solve this problem, a two-step road network extraction framework is proposed. In the first step, the edge information of the road is extracted using a linear detector. To reduce false edges, a false edge removal algorithm based on the directional information of the edges is proposed. In the second step, an improved region growing (RG) algorithm is proposed, which can substantially improve the integrity of the road network extraction compared with the traditional RG algorithm. Finally, the proposed algorithm is validated by GF-3 satellite SAR images.