基于旋转不变性的红外序列图像自动匹配  

Automatic Matching of Infrared Sequence Images Based on Rotation Invariant

作　　者：刘景正[1] 余旭初[1] 刘航冶[1,2] 

机构地区：[1]信息工程大学测绘学院,河南郑州450052 [2]江西省数字国土重点实验室

出　　处：《红外》2009年第9期35-40,共6页Infrared

基　　金：江西省数字国土重点实验室开放研究基金资助项目(DLLJ200901);国家863项目(2006AA12Z154)

摘　　要：红外序列图像的匹配处理是在一系列红外图像内寻求像素点对应关系的过程,它直接涉及并影响到红外图像配准的精度,是视频目标识别、定位、跟踪等处理的关键技术之一。匹配中为了加快搜索速度,采用了金字塔策略。针对红外图像信噪比低的特点,预处理时引入了小波阈值算法对红外图像进行滤波处理;针对无人机红外序列图像间旋转角度大的特点,采用SIFT算法对灰度序列图像进行了匹配,并结合相关系数控制的方法,进一步去除错配点。匹配时不受传感平台较大角度旋转的影响,从而取得了良好的效果。The matching of infrared sequence images is a process to search the corresponding relation- ship between the pixels in a series of infrared images. Since it can directly affect the registration precision of infrared images, it is one of the key techniques for the data processing, such as the recognition, location and tracking of an object in video images. In order to increase the search speed, the Pyramid Strategy is used in image matching. In view of the features such as low S/N ratio of infrared sequence images, the wavelet threshold algorithm is introduced to filter the infrared images in preprocessing. For the infrared sequence images of Unmanned Aerial Vehicles （UAV）, the SIFT algorithm is used in image sequence matching and is also combined with the correlation coefficient control method so as to get rid of the wrong matching points. Because the matching is not affected by the rotation of the sensing platform at a large angle, good effectiveness is obtained.

关 键 词：红外序列图像 去噪 SIFT 旋转不变性 图像匹配 无人机视频 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]