基于贝叶斯网络的视频监控盲区人群状态推演

Predictions for the Crowd Status at Uncovered Regions by Cameras Using Bayesian Network Model

公共场所人群具有密度高、流动性大等特点,易发生拥挤踩踏等突发公共事件.由于目前对监控摄像机的布设大都离散无重叠,造成现有基于视频分析的人群监控主要面向独立的监控摄像机,难以监测整个区域的人群状态.将视频监控与GIS技术有机结合,设计了监控盲区人群状态的贝叶斯网络推演模型,并对未布设监控摄像机路段人群状态进行了推演.结果表明:利用本文提出的监控盲区人群状态推演方法,可推测人群运动速率、人群流量及人群密度等人群状态的空间分布,基于此可实现对区域人群状态的实时态感知,可为安防部门的密集人群智能管理提供技术支撑.

The massive crowd gathering appears frequently with the rapid urbanization and the development of socioeconomic.Public places with high crowd density have potential dangers to emergencies,such as stampedes.Numerous surveillance cameras have been installed in many cities.However,these cameras have no overlapping surveillance areas in most cases.Existing studies about the crowd monitoring using video are mainly limited to one single camera,which cannot obtain the spatial-temporal patterns for crowd status at large scales.In this study,we proposed a method of prediction for the crowd status in uncovered parts of the road network by cameras using Bayesian network model with the integration of video surveillance systems and GIS.Experimental results show that we can obtain the spatial-temporal information for the crowd status at large scales using the proposed method in this study,including crowd movement velocity,crowd flow,and crowd density.All the spatial-temporal information of crowd status can provide scientific basis for the intelligent management of the high density crowd in public places.