The Application of Improved SIFT Algorithm in High Resolution SAR Image Matching in Mountain Areas

Because of the influences of imaging characteristic and the terrain of survey area, SAR images acquired have server geometric and radiometric distortion, particularly evident in mountain areas. In addition, SAR images have heavy speckle noise usually. Therefore, traditional image matching methods have some limitations for SAR image matching. The paper proposed an improved SIFT matching algorithm. The paper used TerraSAR-X and Radarsat-2 images from the same survey area in Hengduan Mountain in western China as experimental data This survey area has complex landform types, high altitude, and big elevation difference. We carried out feature points extraction and automatically matching experiments with the homologous and heterologous SAR images. For homologous SAR images, the accuracy of matching could be up to 99%. While for heterologous SAR images, it was about 86%. Experiments show that the improved SIFT algorithm is stable and reliable, has good robustness, effectiveness and suitability.