几种卫星合成孔径雷达影像配准算法的比较研究

Comparison of algorithms for co-registration of satellite synthetic aperture radar images

复数影像配准是卫星合成孔径雷达干涉数据处理中的重要步骤之一,影像配准精度将直接影响干涉条纹图的质量和由此提取的高程或位移结果的精度。文中利用上海地区ERS-1/2影像数据进行实验,对比分析了三种方法的配准质量和运算效率,试验结果表明:在配准质量方面,相干系数法最高,相位差影像平均波动函数法次之;而在信噪比较低的情况下,最大频谱法的配准质量优于相位差影像平均波动函数法。在运算效率方面,相位差影像平均波动函数法的运算效率最高。

The co-registration of complex images is one of the vital procedures in data reduction of satellite synthetic aperture radar (SAR) interferometry that can be applied to extract either ground elevations or displacements.The imperfect coregistration can degrade the quality of interferogram fringes,and thus lowering the accuracy of elevations or displacements derived interferometrically.By using a pair of SAR images collected over Shanghai through the tandem mode of ERS1 and ERS2, this paper carries out computations using three co-registration algorithms to compare their performances on quality and computation cost in co-registration.Some findings have been obtained from the experiments. It is found that CCM has the best performance in terms of coregistration quality, and MAF ranks the second. However, MSM is superior to MAF when the signal noise ratio (SNR) is low in the interferometric dataset. In terms of computational efficiency, MAF takes the least time.