A novel change detection method based on high-resolution SAR images for river course

Monitoring changes of river courses by using SAR images is an important part of land-cover change detection. Due to difficulties in change detection of river courses, which are induced by the high spatial-resolution of SAR images, we propose a novel change detection method in this paper. Firstly, we accurately extract river courses from each single-temporal SAR image. To address this issue, we propose strategies based on contourlet transform for speckle noise reduction, marker extraction from river courses at the best decomposition scale, marker-based watershed segmentation and region merging. Then, we filter out the false targets which are similar to the river courses from the segmentation results. Finally, through directly comparing the difference among river courses in multi-temporal images based on registration results, we achieve the final detection of river course changes. In the experiments, we use multi-temporal high-resolution SAR images from different sensors, compare and analyze the experimental results with other methods, which confirms the reliability and effectiveness of the proposed method.