多传感器图像的二次匹配算法

针对光学图像的多光谱、多传感器图像的自动配准,提出了一种基于边缘重心的多传感器图像二次匹配算法.首先,通过边缘检测和相应的后期链接处理,提取出较为完整的图像边缘;然后,根据图像边缘的链码相关以及区域不变矩策略,寻找匹配边缘,并对已匹配边缘的重心进行一致性检验获得控制点对;最后,采用最小二乘法估计图像粗匹配变换参数.在粗匹配的基础上,根据匹配后图像共有区域内边缘的主要方向和重心,再次对图像进行细匹配,修正匹配参数以提高匹配精度.多种遥感图像数据的配准实验和对比试验证实了该自动算法具有较高的可靠性和配准精度.

A two-step registration algorithm for multi-sensor images

Considering automatic registration of multi-spectrum, multi-sensor remote images, we put forward a two-step image registration algorithm based on edge center. First, through edge detection and corresponding interlinkage processing, long boundary regions am extracted. Then, the correspondences of long boundary re- gions am developed by chain code correlation and invariant moments. And the center of gravity in the correspondent regions is used as primitive points, and least square method is used for estimating the transformation parameters of coarse image matching. Finally, according to the direction and the gravity center of short boundary regions of the image, a fine registration is carried out for it, and the parameters am modified to improve the accuracy of registration. Experimental results have verified high accuracy and reliability of this algorithm.