中文摘要：

合成孔径雷达SAR用于图像匹配导航系统,可大大提高导航精度。然而地形起伏和大气干扰等因素使得SAR实时图存在几何失真,降低了匹配性能。通过对SAR实时图几何失真成因分析,提出了一种二维可分离的几何校正方法,并得出方位向几何失真可跟随成像处理预先校正,图像后处理只需校正距离向。利用数字高程模型DEM和正射图模拟得到了3幅具有不同地形特征的SAR实时图,并分别利用3种几何校正方法对其进行统计实验,得到了不同地形SAR实时图几何校正方法选取的准则。

英文摘要：

The accuracy of scene matching navigation system can be improved greatly when synthetic aperture radar （SAR） is used. However, geometric distortions of SAR real-time image, which are caused by terrain undulation and atmosphere turbulence etc, decrease the matching performance. By analyzing the causations of geometric distortion, a two-dimension separable correction method is presented, which corrects the distortion in azimuth direction by phase compensation along with the imaging processing, and leaves the distortion in range direction to the image post processing. In the end, an experiment was conducted based on three SAR real-time images of different terrains. Each of the images is simulated by digital elevation model （DEM） and orthography, and then corrected by three different methods. Based on the experimental results, some rules of the correction method selection are acquired.