Road Network Extraction from SAR Images Based on GaoFen-3 Satellite Image

Road is an important part of the transportation network, and with the acceleration of urban and rural construction, how to quickly identify and extract the roads and update the road network becomes a crucial issue. Due to the coherent imaging mechanism, the SAR images of GaoFen-3 are inevitably speckled, which seriously affects the interpretation of the images. Especially the detection of dark and weak roads under strong speckle interference is difficult. In the actual production, most of the enterprises still extract the road network from SAR images by traditional manual outlining, which has a large workload and high labor cost. Therefore, we propose a method to extract roads from GaoFen-3 SAR images using deep learning, and the results of testing on data from two different regions show that the method has good extraction accuracy and has the potential to be applied to large-area extraction of roads from GaoFen-3 images.