基于不变特征的CCD影像与红外影像的自动配准  

Automatic registration of CCD images and IR images based on invariant feature

作　　者：丁琳[1] 倪希亮[1] 江涛[2] 胡顺石[1] 

机构地区：[1]中国科学院遥感应用研究所,北京100101 [2]山东科技大学测绘学院,山东青岛266510

出　　处：《红外与激光工程》2011年第2期350-354,共5页Infrared and Laser Engineering

基　　金：国家科技支撑计划项目(2007BAH15B01);国家重点基础研究发展规划项目(2007CB714406);国家航天局航天遥感论证中心资助项目(O7K00110KJ)

摘　　要：根据环境星可见光影像和红外影像的成像特点,针对大幅影像配准时计算量大、配准点分布严重不均匀以及错配率高等问题,提出了一种基于尺度不变特征变换(SIFT)的改进配准算法。首先,通过对影像进行分块提取SIFT特征关键点,实现了影像特征点的均匀化,同时加快了特征点的提取速度;然后,局部自适应地对特征点进行几何约束,从局部选取最优匹配点以达到提高图像匹配准确率的目的;最后,通过一致性检测原理剔除错配,实现最终的精确配准。利用所提出的算法,通过对环境星可见光影像与红外不同分辨率影像进行大量配准实验,结果表明:该算法能够快速、准确地实现环境星可见光影像与红外影像之间的配准,具有很好的实用价值。According to the imaging mechanism of IR and visible images from the environment, the problems of highly computing complexity, serious non-uniform distribution of matching points and high wrong registration rate for large images were studied, then an improved image registration approach was presented based on SIFT. First, the improved SIFT algorithm realized the uniform distribution of SIFT feature points by image segmentation, and also speeded up the extraction rate of SIFT feature points. Then, the best block matching points were selected adaptively using geometric constraint to improve the image matching accuracy. Finally, some wrongly matched points were eliminated by using consistency checking, which further improved matching accuracy. A large number of experiments from HJ-1 visible images and different resolution IR images show that the proposed algorithm can quickly achieve accuracy registration between HJ-1 CCD images and infrared images, and has a good practical application.

关 键 词：自动配准 SIFT 红外影像 均匀化 

分 类 号：TP722[自动化与计算机技术—检测技术与自动化装置]