星载合成孔径雷达构像方程及其应用于轨道精化的研究

Imaging Model for Space-borne SAR and its Application in Orbit Refinement

星载合成孔径雷达构像方程是地理编码、立体测量、影像定位等几何处理的基础。本文首先讨论了构像方程的参数选择,在充分考虑轨道物理模型及摄动力影响的基础上,提出了一套改进的轨道模型参数。从距离方程及多普勒方程出发,详细推导了包含轨道模型参数及成像处理参数的星载SAR构像方程。作为构像方程的一个具体应用,本文研究了星载SAR影像的轨道精化问题。将构像方程线性化,对每个地面控制点列出误差方程,可以用类似于航空影像的空间后方交会的方式精确求解轨道模型参数。试验中从1∶5万地形图上量取了5个地面控制点,对一景Radarsat SAR影像进行了轨道精化。比较发现,头文件提供的轨道与精化的轨道之间有约2km的位置差异,用精化的轨道数据得到了高精度的几何校正结果。它间接地表明了轨道精化算法的正确可行。

Imaging model is the basis for most geometric processing of space-borne SAR imagery, including image rectification, radargrammetry and image geolocation. This paper develops a parametric imaging model for SAR, which explicitly describes the relationship between the 3D cartographic coordinate of a ground target and the 2D image coordinate of the corresponding pixel. The parameters of imaging model consist of near range, image starting time, Doppler parameters and orbit parameters. After quantitatively evaluating the influence of the Earth perturbations on satellite orbit, we present an improved satellite orbit model, which uses less parameters and is able to fit the space orbit at nearly the same accuracy compared with classical model. As an application of the imaging model, an orbit refinement algorithm using ground control points （GCPs） is presented. For this purpose, the imaging model is rewritten to incorporate only 9 orbit parameters, and then linearized through differentiating with respect to the orbit parameters. For each GCP, the image coordinates are expressed as the function of the orbit parameters and two error equations are formed. The orbit parameters can then be iteratively determined through solving a linear equation set consisting of the error equations from all GCPs under least square rule. The orbit refinement algorithm is tested on a RADARSAT SGF （SAR Georeferenced Fine resolution product） scene. 5 GCPs are identified from 1：50000 scale topographic maps to refine the orbit parameters coming with the SGF image. The accuracy of the refined orbit is demonstrated through geometric correction of the SGF image. While the SGF data is poorly corrected using the original orbit data, it is precisely corrected with refined orbit parameters.