摘要
在分布式摄像机视频监控中,传统单独使用PHD(部分Hausdorff距离)等方法进行运动目标匹配效率低下;对此,提出了一种高精度的图像匹配算法。首先采用混合高斯模型从各个摄像机视频中提取运动目标,提取目标的harris角点,采用NCC(归一化互相关)初步确定角点中的匹配点对,采用PHD做二次匹配,去除明显误匹配的点对,最后结合基础矩阵计算匹配点对的匹配度。实验表明,该方法提高了匹配效率,且具有较高的匹配精度。
For moving objects matching of distributed cameras in modem video surveillance, the traditional PHD method is inefficient. For this, a high-precision image matching algorithm is presented. First, moving objects from each camera video by Gaussian mixture model is extracted and got the Harris comers of the objects. Then, determine the matching comers using NCC algorithm and PHD for twice-matching as a result of which the obvious false matching points are removed. Finally, the matching rate through the fundamental matrix is got and the marching accuracy is improved on.
出处
《科学技术与工程》
北大核心
2014年第21期267-270,共4页
Science Technology and Engineering
