视频中行人快速检测计数方法研究

为了解决智能监控中对行人在实时帧中检测并计数的问题,提出一种快速高效的检测计数方法。采用高斯背景建模和边缘检测获取运动区域及其轮廓信息,形成检测区域;同时根据头部和肩部的组合形态构建行人检测模版,最后对检测区域进行模版匹配,统计行人数目。实验结果表明,该方法快速高效,有较高的正确率和较低的统计误差,同时对行人遮挡有一定的区分能力,对阴影也有较强的抗干扰能力。

In order to solve pedestrian detection and counting in real-time frames at intelligent monitoring,a fast and efficient detection and counting method is proposed in this paper. Using Gaussian background modehng and edge detection method to obtain the motion objects and their edge information as detection areas; on the other hand a pedestrian detection model is constructed by the head and shoulder combination form.Finally,use the model to do matching on the detection areas and count the number of the pedestrians.The experimental results show that this method is fast and efficient with a high accuracy and low statistical errors,and it can detection pedestrian well when there are overlap or shadow.