在总结现有算法的基础上,提出了基于相干系数、Harris特征点、小波金字塔及TIN三角微分纠正技术的单视复数雷达图像的配准流程。通过ERS 1/2的实验表明,提高了配准的精度和效率,特别是保证了在没有精确轨道甚至没有轨道参数的情况下也能获得很高的配准精度,计算正确的干涉相位图。在重采样过程中采用的三角联网策略,进一步使匹配点的局部拟合误差得到有效控制,得到配准精度更高的复图像对。在CPU 3.06GHz计算机上,43 s内完成5000像素×1000像素的主辅图像的配准,平均相干系数为0.719855。

This paper mainly discusses the high-precision co-registration methods for interferometric synthetic aperture radar(InSAR) complex image.After several registration methods are reviewed and compared,a new method of the co-registration for two single looking complex(SLC) SAR images(the master image and the slave image) based on coherence coefficient,harris feature points, wavelet pyramid and Triangulated Irregular Network(TIN) based differential rectification was proposed.The experiment with space-borne InSAR(ERS 1/2) showed the method accomplished co-registration in about 43 seconds with a coherence 0.719855.It proves that the accuracy of this method can satisfy the requirement of Interferometry.