With the improvementof the intelligence degree of video surveillance system, the head counting in surveillance area becomes a studying focus too. This paper proposes a fast and effective head counting method. In this method we use the combination of background subtraction and frame subtraction to extract the motion objects first, and then use coarse-to-fine segmentation to distinguish the group people and the single passer-by, at last an algorithm based on object area is used to count the number of people groups. Compared with existing methods,the proposed method has got great improvement in real-time when keeping high accuracy unchanged ,and reduces the requirements on hardware as well.%随着视频监控系统智能化程度的提高,监控区域的人数统计也成为一个热点.提出了一种快速有效的人数统计方法.使用背景差分与帧差分相结合的方式提取运动目标,使用两步分割法区分人群与单个行人,使用基于目标区域算法统计人群的数量的.与目前已有的方法相比,该方法在保持较高准确率的前提下,实时性有了很大的提高,并且降低了对硬件的需求.
展开▼