动态视觉监控系统人流量检测  

Dynamic visual pedestrians flow measurement system

作　　者：贾其臣 张琦[1] 曹开发[1] 谢慕君[1] 

机构地区：[1]长春工业大学电气与电子工程学院,吉林长春130012

出　　处：《长春工业大学学报》2017年第1期79-83,共5页Journal of Changchun University of Technology

基　　金：吉林省科技厅基金资助项目(2015-25)

摘　　要：运用混合高斯模型对背景进行建模,然后实时更新背景,检测出运动目标,采用改进卡尔曼滤波对目标位置进行预测,并与目标实际位置进行参数匹配,如果二者距离不大于阀值,则认为二者为同一目标,否则为二个不同目标。视频图像的人数统计实验验证了检测方法的有效性。Mixed Gaussian model is applied to build and update the background to detect the moving targets. With improved Kalman filter, target positions are estimated and then matched with the real position parameters. If the distance between the estimated and real is less than the threshold, we can conclude that the two positions are thesame object otherwise they are different. Statistic experiments of pedestrian flow based on video images verified the method.

关 键 词：人流量检测 混合高斯模型 卡尔曼滤波 目标计数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]