基于相似三角形匹配的红外与可见光图像配准方法  

Method of infrared and visible image registration based on similar triangles matching

作　　者：陈洁[1] 付冬梅[1] 刘燕[2] 

机构地区：[1]北京科技大学信息工程学院,北京100083 [2]装甲兵工程学院装备再制造技术重点实验室,北京100072

出　　处：《激光与红外》2010年第2期215-218,共4页Laser & Infrared

基　　金：国家自然科学基金项目(No.60573016);国家科技支撑计划项目(No.2007BAB27B02-04)资助

摘　　要：提出了一种基于相似三角形匹配的红外与可见光图像配准方法:首先将提取出的Harris角点组成三角形,然后利用本文提出的搜索算法,从红外与可见光图像中找出一对最优的相似三角形,并利用相似三角形的性质,计算出相关的匹配点,最后通过RANSAC算法拟合出配准参数,对图像进行配准。实验结果证明:此算法与传统的基于特征点的图像配准方法相比精度更高,具有较强的鲁棒性,可以解决刚体变换下红外与可见光图像配准的问题。This paper presents a method of infrared and visible image registration based on similar triangles matching. First, Harris comer will be extracted to constitute the triangles. Second, the search algorithm this paper proposed is be used to search the similar triangles which are the best matching and then some related the matching points are be calculated by the nature of the similar triangles. Finally registration parameters are be estimated by RANSAC algorithm. Experiments show that this method can obtain a higher registration accuracy compared with traditional method of image registration based on feature points, has better robustness, and can be used to solve the infrared and visible images registration with rigid transformation.

关 键 词：图像配准 角点 相似 三角形 RANSAC 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]