基于粗DEM的立体SAR像对配准

影像配准是雷达立体摄影测量技术获取数字高程模型的一个关键步骤,也是一大难点。为了能够自动而又稳健地获取高精度的立体SAR像对配准结果,文中设计一种基于粗DEM、距离多普勒SAR构像模型和基于区域逐级窗口的立体SAR像对配准方法。该方法采用距离多普勒SAR成像模型和粗DEM获取像素级的粗略配准结果。在此基础上,采用基于区域的逐级窗口配准方法获取高精度的配准结果。文中最后利用立体SAR像对进行实验研究。实验结果表明,设计的立体像对配准方法可行,其配准精度可以达到子像素级,距离向为0.3像素左右,方位向为0.1像素左右,完全能够满足雷达立体测图的要求,而且该方法还具有很好的稳健性。

Stereo SAR imagery co-registration based on coarse DEM

Due to the new high resolution imagery which can be acquired by TerraSAR‐X ,COSMO‐SkyMed and RADARSAT‐2 sensors in SpotLight mode ,the radargrammetric approach is exploited to generate a digital elevation model by more and more researchers from these high resolution SAR imagery .Image co‐registration is the major step in the radar stereo process to acquire digital elevation model . Image registration is a vital challenging task for stereo SAR technique .A fully automatic registration approach w hich is accurate ,robust ,and fast ,is required .For this purpose ,a new method that fuses the RD model and area‐based technique is provided .The coarse registration result is acquired with RD model and coarse DEM .Next ,the area‐based process is implemented by the maximization of normalized cross correlation (NCC) .A Terra‐SAR‐X stereo pair is used to test the effectiveness of the proposed algorithm .The result shows that the new registration method is feasible ,and registration accuracy is up to subpixel with 0 .3 pixel in range and 0 .1 pixel in azimuth . T he registration result can satisfy the requirement of radargrammetry to generate DEM .The new method is also robust .