一种基于 RPC 模型的 SAR 影像几何精纠正新方法

摘要

Synthetic Aperture Radar (SAR ) images have advantages of being usefulness all-weather and all-phase. Therefore,they have been widely used in terrain mapping and change detection.Rational Polynomial Coefficient (RPC)model is a novel method being used in SAR image geo-location,which is a highly accurate general model comparing with the physical model and can be used to take place of the physical model.This paper utilizes GCPs (Ground Controlling Points)to achieve RPC model refining and puts forward a novel method of SAR geocorrection based on the refined RPC Model.The test site is in Genhe district in Inner Mongolia province in China and the test images are TerraSAR-X and Radarsat-2 images.This paper concludes that:(1)RPC Model has a highly accuracy compared with R-D model,which is all most 10e-6 pixel.(2)Based on the refined RPC model,the geo-location accuracy of TerraSAR-X image is 0.01 pixel and the one of Rasarsat-2 image is 0.1 pixel. (3)The quality of orthoimages of geocorrection can be several meters based on refined RPC model.%RPC（Rational Polynomial Coefficient）是最新提出的应用于 SAR （Synthetic Aperture Radar）影像定位的模型算法，该模型通过对严格物理模型的高精度拟合，达到替代严格物理模型的目的。本文在 RPC 模型的基础上，利用控制点对 RPC 模型进行精化，提出了一种基于 RPC 精化模型的 SAR 影像几何精纠正新算法；采用内蒙古根河地区的 TerraSAR-X 卫星影像和 Radarsat-2影像进行几何精纠正实验，验证了算法的有效性。实验结果表明：①RPC 模型能够实现对 R-D 模型的高精度拟合，拟合精度在10e-5像元左右；②使用精化后的 RPC 模型， TerraSAR 影像的定位精度达到百分之一像素，Radarsat-2影像的定位精度能够达到十分之一像素；③基于 RPC模型的 SAR 影像精几何精纠正算法生成的正射影像精度达到米级。