视频监控中的行人检测与计数

Pedestrian detection and counting in video surveillance

采用简化的混合高斯模型提取背景并保持更新,利用背景差法提取前景目标实现行人的检测。根据运动目标在相邻两帧之间的运动不会发生突变这一事实,采用Kalman滤波器进行预测估计,提出一种简单快速的方法对行人进行跟踪匹配。在跟踪的基础上,设置了两条检测线统计过往的行人数目。经过实验证明,我们设计的系统能够满足实时要求。

The simplified mixed Gaussian model was used to get the background and keep updating the background,and then the background difference method was used to get the foreground target and detect the pedestrian.According to the fact that the movement between two frame pictures can not mutate,using Kalman filter to predict the trajectories,we proposed a simple and fast method to track the pedestrian.Based on the motion tracking,we set two detected lines to count the number of people passing by.The experiment results show that our system could satisfy the request of real time with high accuracy.