红外与可见光复合寻的制导中的快速图像配准方法  

A fast image registration algorithm for compound IR/visible light homing guidance

作　　者：李露[1] 蒋宏[2] 任章[2] 

机构地区：[1]北京航空航天大学宇航学院,北京100191 [2]北京航空航天大学自动化学院,北京100191

出　　处：《哈尔滨工程大学学报》2009年第2期228-232,共5页Journal of Harbin Engineering University

摘　　要：对现有多传感器图像配准方法在红外与可见光复合寻的制导应用中存在的计算量大等不足,提出了一种基于部分Hausdorff距离的快速方法和一种两集合相似度的定义,从而实现了有效减小误匹配特征点对的红外与可见光图像快速配准方法.此算法通过分别提取两幅图像边缘上显著特征点,再根据提出的快速方法和两集合相似度定义,粗略估计出图像间的旋转角度、平移量;然后依据此估计值,缩小特征点匹配时匹配特征点搜索范围,求得一致特征点对;最后用最小二乘法求解最优变换参数.实验结果表明了该方法的有效性、快速性.To overcome the limitations of large computational cost in current multi-sensor image registration algo- rithms for compound IlL/visible homing guidance, the authors proposed a fast method based on partial Hausdorff distance and a definition of similarity measurement for two groups. This method has realized a fast IR/visible image registration algorithm that effectively decreases mis-matching of feature point pairs. First, this algorithm extracts re- markable feature points from the image edges; second, based on the proposed fast method and the similarity meas- urement definition, it roughly estimates rotational and translation parameters; thirdly, according to the estimated parameters, it nam）ws the search scope in matching feature points and choosing the matching feature pairs; finally, it calculates the optimum transformation parameters by the least square method. Simulation results indicate that this registration algorithm is not only effective, but also fast.

关 键 词：图像配准 边缘提取 方向方差 部分Hausdorff距离 交叉验证 仿射变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]