一种基于重叠子孔径回波信息的SAR图像配准算法

SAR image registration algorithm based on echo information of overlapping subaperture

合成孔径雷达(SAR)图像配准是寻找多幅SAR图像之间的几何变换关系的过程,使不同图像校正到统一空间坐标系。传统光学图像配准方法,如尺度不变特征变换(SIFT)结合随机抽样一致(RANSAC)用于SAR图像配准时,由于受到乘性相干斑噪声的干扰,配准性能受到较大影响。提出一种基于重叠子孔径回波信息的SAR图像配准算法,在对SAR回波复数据进行成像处理的过程中,利用相位信息并结合子孔径自聚焦方法得到强相关的重叠子孔径图像,并利用多图像配准方法实现成像孔径内和孔径间的子孔径图像配准,提高图像配准精度。多个不同场景的实测数据处理结果表明:所提算法相比于SIFT+RANSAC算法,同名点数量增加,误匹配点对减少,均方根误差减小,对相干斑噪声具有较强的鲁棒性,配准效果得到了显著提高。

Synthetic Aperture Radar(SAR)image registration is the process of finding the geometric transformation relationship between multiple SAR images to calibrate different images to a unified spatial coordinate system.When traditional optical image registration methods,such as Scale-Invariant Feature Transform(SIFT)combined with Ran⁃dom Sample Consensus(RANSAC)algorithm,are used for SAR image registration,the registration performance is greatly affected by the interference of multiplicative speckle noise.In this paper,a SAR image registration algorithm is proposed based on the echo information of overlapping subaperture.In the imaging processing of the complex echo data,using the phase information and the subaperture autofocusing algorithm,highly correlated overlapping subaper⁃ture images are obtained.The multiimage registration methods are used to achieve the subaperture image registration within and between imaging apertures,so as to improve the accuracy of SAR image registration.The processing re⁃sults of the measured data of several different scenes show that the proposed algorithm has more homonymous points,fewer mismatched point pairs,and smaller root mean square error than the SIFT+RANSAC algorithm.More⁃over,the proposed algorithm is more robust to speckle noise,and thereby gains the improved registration performance.