SAR Simulation Based Matching of UAV Images to Very High-Resolution SAR Images

Our experiment focused on the simulation of high spatial resolution SAR for geometrical matching between very high-resolution UAV images and geometrically precise SAR images. To achieve this, UAV images of the experimental area were acquired, and a three-dimensional photogrammetric model was generated from the UAV data. This photogrammetric model provides source information for the SAR simulation. The following matching experiment is based on several feature matching algorithms, specifically Scale-Invariant Feature Transform (SIFT), Oriented FAST and Rotated BRIEF (ORB), which were used for the co-registration of simulated and real SAR images. The effectiveness of these algorithms was quantified using the matching rate as a key evaluation metric. The experimental results demonstrate that ORB outperforms SIFT primarily owing to its enhanced robustness against rotational variations and noise. Consequently, ORB is established as a versatile and superior choice for high-resolution SAR image registration tasks and is applicable to both real and simulated data.