基于结构的SAR图像配准

Structure-driven SAR Image Registration

由于SAR图像中相干斑的存在,使得已有用于光学遥感图像自动配准的算法往往无法直接应用。基于人工通过地物结构推理匹配关系进行配准的想法,提出了一种基于结构的SAR图像自动配准算法。该算法首先通过检测出的点目标构造“虚拟结构”,然后再综合新提出的“虚拟结构”不变量及结构区域不变矩作为相似测度完成匹配检测,最后用LMS算法估计出变换参数从而实现SAR图像的自动配准。实验结果表明,该算法不仅能够有效实现SAR图像的自动配准,而且能有效避免SAR图像中相干斑对配准过程中特征检测和匹配造成的影响。

Due to the presence of speckle in SAR image, the existing registration algorithms which are successfully used in optical remote sensing image are usually not applicable to SAR image directly. Based on the idea that human observers could greatly simplify the matching problem at the level of object structure, a structure-driven SAR image automatic registration algorithm was proposed. Firstly, the virtual structures were constructed from the point targets detected from the SAR images; and then they were matched by using the combination of structure invariance and structure region invariant moments.Finally, the LMS algorithm was used to estimate the affine transformation parameters, thus the SAR images could be registered automatically. Practical experimental results show that the proposed new algorithm is not only valid in the automatic registration of SAR images, but also can avoid the influence caused by speckle in feature detection and feature matching process.