基于视频分析的高密度人群踩踏事故风险防控方法研究

A stampede in a crowded place takes on the characteristics of sudden occurrence, difficult control and multiple induced reasons; and usually leads to severe casualties. It has scientific value and practical significance to study the risk prevention and control methods for stampede in highly dense crowd. In this project, an extraction algorithm for crowd motion characteristics in large-scale scene and with high density is developed, and its change mechanism and the corresponding critical condition are studied. A pedestrian dynamical model representing crowd motion characteristics is established. Based on the above research results, the space-time characteristics of real-time risk and its critical value for stampede are studied, and the risk factors of crowd activities are analyzed and the risk values are quantitated. With the risk evaluation results, the emergency plan guide for stampede in highly dense crowd based on the scenes developed by the pedestrian dynamical model is presented, and the safety scheme guide for crowd activities is provided. The research outcomes will provide theoretical basis to decrease stampede risk, and reduce the injuries and deaths in crowd activities; thus provide scientific support to enhance the emergency level for crowd places.

人群密集场所的踩踏事故具有发生突然、难以控制、诱发原因多等特点，一旦发生将造成严重的人员伤亡，研究高密度人群踩踏事故风险防控方法具有重要的科学价值和现实意义。本项目采用视频分析方法发展大场景下高密度人群运动特征的提取算法，并研究其运动特征的转变机理及临界条件，建立体现高密度人群运动特征的行人流模型；在此基础上，研究高密度人群运动过程的实时风险时空变化特征及人群踩踏事故的风险临界值，分析人群聚集活动的风险因素并量化其风险值；以风险评估结果为依据，提出基于行人流模型构建的踩踏事故情景的高密度人群踩踏事故的应急预案编制指南，及人群聚集活动安全方案指南。为降低人群聚集活动中人群踩踏事故风险，减少人员伤亡提供理论依据。进而为我国提高人群密集场所应急管理保障水平提供科学支撑。

针对人员密集场所踩踏事故风险防控的前沿问题和主要难点，本项目提出了高密度人群运动特征的提取算法，建立了行人流模型，提出了人群运动过程实时风险时空分布的计算方法，搭建了以人群聚集风险时空分布为核心的人群聚集风险预警系统，为城市人群密集场所的风险监测预警提出科学手段。项目成果共发表SCI论文9篇，做7次大会特邀/邀请报告，获北京市科学技术奖二等奖；项目负责人连续3年入选中国高被引学者榜单，并获国家杰出青年科学基金项目资助。项目成果为降低人群聚集活动中人群踩踏事故风险，减少人员伤亡提供理论依据，进而为我国提高人群密集场所应急管理保障水平提供科学支撑。