Study On Method Of Matching For Synthetic Aperture Radar Images

Image matching is one of the important image processing steps of Synthetic Aperture Radar (SAR) image, especially for that of SAR stereo-photogrammetry. The method and model of SAR image matching is the key of research in SAR data processing recently. Due to the different imaging mode and mechanism from optical sensor, there is a large difference between SAR image and optical image. This research started from analyzing the imaging geometrical mode of SAR, combined traditional optical method with SAR imaging model, tried to find the new matching model which could be suitable for SAR image.Based on the traditional method of optical epipolar geometry, the SAR image matching method using corresponding epipolar line was presented. Firstly the relationship between images pair was built, secondly the similar epipolar line image was generated by two methods based on both coplanarity condition and epipolar definition, finally both same-side and opposite-side stereo SAR image pairs were tested by these methods. It was proved that the classical optical method could be applied in SAR image, and the similar epipolar line SAR image could be generated by means of SAR imaging equation and epipolar definition. But this method of gray level correlation couldn't get the better result.Considering of SAR own image characteristic and imaging mechanism, the method combined feature and gray level was studied in the paper, which was on the foundation of similar epipolar line SAR image. After the different operators'effect in SAR edge detection was analyzed, combination operator of Canny and Ratio was adopted to do the edge detection, and the feature points were used to finish the SAR image matching. The test result of this method was better, and the efficient of SAR image matching was also improved.By the result of the research and tests, it was proved that the SAR image matching methods of based on both epipolar geometry and combined edge feature and gray level were feasible, and this work provided the technical foundation for stereo-photogrammetry for SAR images.