Research on SAR Image Matching Based on SAR-Harris Detector and Adaptive Locally-Affine Matching

Synthetic aperture radar (SAR) image matching is an essential step in SAR image processing. The synthetic aperture radar-scale invariant feature transform (SAR-SIFT), which combines the ratio of exponentially weighted averages (ROEWA) operator with multiscale Harris detection and employs Random Sample Consensus (RANSAC) for matching and filtering, has shown promising results. However, SAR-SIFT suffers from low matching efficiency in specific scenarios, and RANSAC exhibits strong randomness and potential problems of non-convergence. To address these problems, this paper introduces an approach that combines scale-space layers reduction in the SAR-Harris detector with the efficient Adaptive Locally-Affine Matching (AdaLAM) to enhance SAR image matching performance. The proposed SAR-SIFT-AdaLAM method is experimentally evaluated on two sets of different SAR images for matching, and the results indicate an improvement of approximately 25% in the number of matching points compared to the SAR-SIFT, along with a reduction of about 30% in matching time, demonstrating its effectiveness.