基于多目标跟踪算法的车站人流聚集监测研究

Research on the Crowd Gathering Monitoring of Station Based on Multi-Object Tracking Algorithm

车站运营过程中实时获取人流信息,对聚集行为与重点区域进行监测、预警、分析在车站安全保障与管理中具有重要意义。目前车站的监控硬件设备数量不断增加、配备复杂,单独依靠监控系统的预警功能实现车站实时、全方位、多角度的人流密度与聚集分析存在困难。为此,提出了一种基于YOLOv5+DeepSort并结合HRNet模型的方法,实现了人员速度、方向、密度、异常聚集、异常侵入、隔栏递物和逃票检测。在车站实时监控数据集和收集的相关互联网数据集上进行了实验,实验结果显示,提出的算法能够实现检测任务,具有实时高效分析监控视频与实际应用的能力。在此基础上开发了一套车站视频行为分析系统后端服务与前端管理平台,对车站旅客出行保障与安全运营具有重要意义,有发展前景。

During the operation of the station,it is of great significance to obtain real-time passenger flow information,monitor,warn and analyze the gathering behavior and key areas in the safety assurance and management of the station.At present,the number of monitoring hardware equipment in the station is increasing and the equipment is complex.It is difficult to achieve real-time,all-round and multi angle passenger flow density and aggregation analysis of the station solely relying on the early warning function of the monitoring system.To solve this problem,this paper proposes a method based on the combination of YOLOv5+DeepSort and HRNet model,which realizes the detection of personnel speed,direction,density,anomaly aggregation,anomaly intrusion,barrier delivery and ticket evasion detection.We have carried out experiments on the real-time monitoring data set of the station and the relevant internet data set collected.The experimental results show that the algorithm proposed in this paper can achieve the detection task,and has the ability to analyze the monitoring video and practical applications in a real-time and efficient manner.On this basis,we have developed a set of back-end service and front-end management platform of the station video behavior analysis system,which is of great significance and development prospects for passenger travel security and safe operation of the station.