摘要
针对可见光与红外图像差异较大导致的匹配困难的实际问题,提出了一种基于蒙特卡罗估计改进Hausdorff距离(MCM-HD)的景象匹配方法。该方法在MCHD的基础上,使用蒙特卡罗方法来估计改进的Hausdorff距离(M-HD),并定义了MCM-HD,即采用随机抽样的特征点子集来计算M-HD,从而有效地减少了计算量。为了提高匹配精度,采用分层MCM-HD与Nprod相结合的方法,在求出距离最小k个点之后采用Nprod相似性度量得出最终匹配位置。与MCHD算法相比,该算法有效提高了匹配精度,同时缩短了匹配时间。
At present, optical and infrared image have large gray value differences between them that will cause big error in scene matching. In this paper, a new method based on MCM-HD for this problem was proposed. The method used Monte Carlo to evaluate the modified Hausdorff distance （M-HD）, and gave the definition of MCM-HD using a randomly sampled set of feature points to evaluate the MHD. As a result, calculation amount of the M-HD was decreased. In order to improve the matching precision, combining the layered MCM-HD with Nprod, calculated the minimum k values, and used Nprod to obtain the more accurate position. Compared with MCHD algorithm, the method effectively improves the precision and shortens the matching time.
出处
《红外与激光工程》
EI
CSCD
北大核心
2008年第2期289-291,共3页
Infrared and Laser Engineering
基金
国防“十一五”预研项目(513220208)
