摘要
本文讨论一种实用化的视频客流计数算法。算法首先生成双行时空图,然后用行帧差提取运动目标区域,利用头部特征提取出运动目标,最后采用区域质心匹配法对人体目标进行匹配与计数。实验证明该算法简单可靠,能够克服光线的变化及干扰物体的影响,并能很好的满足实时性的要求。
The paper discusses a practical video-based algorithm for passenger counting. Firstly, double row space-time images are acquired. Then, in order to obtain moving object region, we propose a row frame difference method, which uses head characteristics to locate the moving objects. Finally, we use the method of region centroid matching for people＇s matching and counting. The experiment results show that the algorithm overcomes the influences of various illuminations and disturbed objects, is simple and reliable, and meets the demand of real-time processing.
出处
《微计算机信息》
2009年第27期50-52,共3页
Control & Automation
关键词
行时空图
行帧差
客流计数
区域质心匹配
row space-time image
row frame difference
passenger counting
region centroid matching
