Robust Feature Matching for Slant SAR Image Registration Based on Differentially Constrained RANSAC

This paper proposes a robust feature matching for slant SAR images based on differentially constrained RANSAC. Firstly, we obtain a set of initial matches provided by SAR-SIFT operator. Then, we set a differentially constrained model with strong constraint in the azimuth direction and no constraint in the distance direction. We use the nearest neighbor contrast (NNC) and the distance offset definition (DOD) techniques to eliminate outliers with large distance distortion. Experiments are carried out on Chinese spaceborne and airborne SAR images. The results show that the proposed method has excellent performance in accuracy, distribution and efficiency, and is suitable for matching SAR images covering areas with large terrain fluctuations such as mountainous areas.