基于有理函数模型的星载SAR影像几何校正

Spaceborne SAR image geocoding with RFM model

主要研究了面向星载合成孔径雷达SAR(Synthetic Aperture Radar)影像几何处理的有理函数模型RFM(Rational Function Model)的建模、求解和应用方法。首先采用与地形无关的方式,利用严密的距离-多普勒模型(Range Doppler,RD)构建虚拟控制点格网来进行RFM建模,实现了一种RFM模型参数的快速无偏解算新方法,能够取得相对于RD模型很高的拟合精度,在几何定位功能上实现了对RD模型的有效替代,同时大大提高了计算效率。然后在此基础上利用RFM模型实现了星载SAR影像的快速几何校正,为了提高几何校正结果的绝对定位精度,引入少量地面控制点对RFM模型进行了像方改正处理,以消除SAR影像几何定位的系统误差,并利用ENVISAT ASAR数据的实验结果验证了本文方法的有效性。

The methodology of RFM modeling for spaceborne SAR and its application in SAR image geocoding are investigated in this paper. The issues associated with model construction, parameter solution and refinement of RFM model are exten- sively studied. When the rigorous sensor model is known, a virtual control point grid can be established in a terrain-independent way for RFM modeling. A new approach is implemented to obtain unbiased solutions of RFM model parameters at a low compu- tation time cost. The resultant RFM model can be used as a reliable substitute for the Range-Doppler （RD） model for geometric processing of SAR images as the fitting accuracy with reference to the RD model is usually better than 0.01 pixels. However, in general there will be geometric biases in the geocoded SAR images because systematic absolute geolocation errors usually exist in the RD and associated RFM models. To solve this problem, control points are used to derive an additional mathematical transformation to remove the absolute geolocation errors in the RFM model. Experimental results with ENVISAT ASAR data show the effectiveness of the proposed method for RFM model construction.