Automatic estimation of crowd density using texture

This paper considers the role of automatic estimation of crowd density and its importance for the automatic monitoring of areas where crowds are expected to be present. A new technique is proposed which is able to estimate densities ranging from very low to very high concentration of people, which is a difficult problem because in a crowd only parts of people's body appear. The new technique is based on the differences of texture patterns of the images of crowds. Images of low density crowds tend to present coarse textures, while images of dense crowds tend to present fine textures. The image pixels are classified in different texture classes and statistics of such classes are used to estimate the number of people. The texture classification and the estimation of people density are carried out by means of self organising neural networks. Results obtained respectively to the estimation of the number of people in a specific area of Liverpool Street Railway Station in London (UK) are presented.     

基于注意力机制的人群计数方法

为准确预测固定场景中的人群计数,在人群分析领域,采用一种融合注意力机制的卷积神经网络(CNN)进行人群计数,该模块结合空间域注意力和通道域注意力,空间域注意力可以编码整个图像的像素级上下文信息,以更准确地表达像素级别的密度图,而通道域注意力可以在不同的通道中提取更多的区分特征使网络显著表达人群的局部区域,并在多个公开数...     

城市公共场所人群拥挤踩踏事故分析

城市公共场所人群高度聚集、流动性大,紧急状态时常发生群死群伤的拥挤踩踏事故,造成大量人员伤亡和恶劣的社会影响。笔者首先提出拥挤踩踏事故分析的一般方法,并用群集指数来表征人群高度聚集这一参量;通过对近年来国内外人群拥挤踩踏事故的分析,认为该类事故以人群高度聚集为条件,由公众造成又危害公众,具有诱因众多、发生突然、难以控制等特点;强调事故预防和控制重点是建立事故应急机制,加强人群的管理和疏导等;同时从公共场所的性能优化设计、人群素质和群集密度、人群管理控制以及信息交流等方面分析其事故原因及提出相应对策,对预防此类事故发生,减少群集伤亡有重要意义。     

考虑熟悉度与恐慌效应的人群引导疏散仿真算法研究*

为研究紧急情况下人群疏散行为特点,结合势场理论和多出口选择模型,建立考虑引导、熟悉度和恐慌效应的PFT-LMES人群疏散算法。通过搭建有障碍和无障碍2种室内场景,研究不同环境熟悉度以及行人密度下引导作用对人群疏散效果的影响。研究结果表明:引导作用能够提高疏散效率,平衡各出口利用率,引导效果随行人密度增加更加显著;总体疏散时间随引导强度的提高表现为先减小后增大;当接受引导的人数占总数的42%时,2出口各自选择人数与出口通行能力比例基本一致,引导效果达到最优;相对于环境熟悉度较高的人群,不熟悉环境行人的引导效果更为显著。研究结果可为大型场所的紧急高效疏散提供参考。     

城市重点公共区域人群聚集风险的实时定量技术

城市公共区域人群高度聚集且流动性大,紧急状态时易发生群死群伤的拥挤踩踏事故,造成大量人员伤亡和社会负面影响。在FIST模型的基础上,提出以人员密度（D）、人员特性（C）、人与人的相互作用（I）以及人群聚集环境（E）作为表征公共场所人群聚集风险的基本参数,接下来对四个参数进行了相应的技术分析。第一,利用人群监控技术估计人群密度;第二,通过现场监测网络得到的人群压力值来表征人与人相互作用的强度;第三,忽略了个体差异对人群整体的影响;第四,把公共场所中导致事故发生的影响因子归结为综合扰动强度,并建立了相应的数学模型表征了这种强度的大小。最终建立了描述人群聚集风险的DICE模型。同时,给出了人群密度阈值、人群压力阈值以及人群聚集风险的总阈值及其判断标准,整个工作将人群聚集风险实时定量及管理技术推向了实用化的道路。     

人群密度对人群拥挤事故的影响

为了研究公共场所的人群密度与拥挤事故的发生概率.基于连续人群流动模型,从人群密度角度探讨了人群拥挤事故发生的机理.由于不同民族个体生理尺寸的差异,人群最大忍受密度不同,以此作为判断人群拥挤事故的标准,并结合我国情况提出我国人群最大忍受密度为9人/m2.最后模拟了某个拥挤事故场景,用该模型对其进行拥挤事故分析.结果表明,连续人群流动模型可以用于预测拥挤事故的发生,对预防和控制人群拥挤事故具有一定的指导意义.     

面向实时人群动力学分析的深度基本图网络

当前行人疏散实验中基本图计算方法通常是通过对每个行人进行跟踪实现的.但这种跟踪方法难以实现实时人群动力学分析.针对这一问题,提出了深度基本图网络.实验提出的网络框架由两个模块组成,即多尺度递归卷积神经网络(MSR-Net)和光流模块,分别对行人密度和行人速度进行估计.具体来讲,MSR-Net学习了输入图像与行人密度图之...     

基于互相关算法的人员密集场所人群运动速度特征研究

研究人群密集场所的人群运动速度特征可以预测人群的运动趋势,在大型活动组织过程中可以对异常人群运动做出预警,避免过度的拥挤及踩踏事件的发生,保证大型群体性活动的安全顺利开展。利用国内某重要城市核心区公共场所人群运动的视频图像,通过互相关算法提取该场所人群的运动速度,并进一步比较通往景区的四条不同路径上人群运动速度的差异性,分析其人群运动特征。分析结果表明单向通道的人群运动速度较大且运动方向基本与通道的两侧边界平行,而双向通道中由于人群中阻尼效应的影响,人群运动速度的大小和方向都发生了不同程度的改变。分析结果可为核心区管理者进行大型群体性活动的组织、人群疏散与引导提供建议,进而为人群拥挤踩踏事故风险防控、拥挤踩踏事故专项应急预案制定及人群聚集活动安全方案编制提供理论支持。     

The development of a movement-density relationship for people going on four in evacuation

Occupant movement in evacuation models has been simulated and predicted based on a number of variables, including crowd density. This study investigates the relationship between crowd density and occupant crawling movement, as a physical response to environmental conditions in fire. It is an attempt to generate the fundamental speed-density relationship that has been developed for walking movement. This is conducted by examining the impact of population size and exit access width on crowd walking and crawling speeds on a flat surface. The findings of the study suggest that exit access width has a significant impact, whereas occupant configuration (population size) plays less of a factor. The results further demonstrate that there is a significant difference in the crawling speed given the exit access width available, due to the secondary effect of crowd density. The relationship between crowd crawling speed and density is best represented by a quadratic regression model. The study concludes with the need to continuously develop new predictive movement methods, or enhance existing ones in order to cope with the level of detail required to ensure occupant safety and model complexity.     

体育赛场人群疏散过程滞留人数定量模型研究

体育赛场历史事故统计分析表明,出口堵塞为导致人群拥挤踩踏事故发生的主要原因.滞留人群是疏散过程最常见的一种人流形式,同时也是拥挤踩踏事故风险的主要承载体.基于人群流量与人群密度关系建立了体育赛场时间维变量的滞留人数定量模型.体育赛场看台不同宽度出口人群疏散计算结果表明,滞留人数不仅对人群疏散时间有直接影响,而且与事故发生概率之间存在一定的关系.该模型可用于指导体育赛场出口设计,疏散路线选择及应急预案的编制等.     

行李箱携带人群汇流过程的基本图实验研究

行李箱携带者作为公共交通枢纽一种典型的人员类型,研究其对人群运动特征的影响对于公共设施的优化具有重要意义。开展了行李箱携带人群在六种汇流角度下的紧急疏散实验,比较了不同汇流角度对行人运动速度及疏散效率的影响,获取了不同汇流角度和不同测量区域的运动基本图。结果表明:1)在实验设置下,比较不同汇流角度?(出口宽度d=3.00 m)发现,?越小(30°),人群运动速度最大(1.86 m/s)、累计疏散时间最短。随着汇流角度的增大,人群运动速度呈现降低趋势(?=180°时,人群运动速度最小(1.40 m/s)、累计疏散时间最长)。表明汇流角度小的设施,有利于提高行人疏散效率;2)比较不同测量区域的基本图(汇流角度一定)发现,汇流交汇处的人群密度分布要小于离交汇处较远的测量区域;3)比较不同汇流角度的基本图(测量区域一定)发现,汇流交汇处的行人密度与流量主要分布在1.34 m-2~2.15 m-2与2.00 (m·s)-1~3.50 (m·s)-1之间。并且,在实验观测的密度范围内,行人流量随着人群密度的增加而上升。研究结果可为优化建筑设施设计和制定人群管理策略提供 依据。     

基于视频分析的人群密集场所客流监控预警研究

为实现人群密集场所客流安全隐患早发现,辅助管理人员早决策,人群聚集风险区早疏散,提升对灾难的预见性和主动性。在国内外人群异常聚集监测预警现状分析基础上,对比分析得出监控视频分析技术是解决人群密集场所精准预警难题较为理想的解决方案;构建以视频智能分析的人群计数、密度估计、行人追踪、活动烈度识别为核心技术的人群密集场所风险预警技术框架;将该技术框架应用到某大型商圈的商业街区,获得监控区域内的人群总数、密度分布、行人轨迹和异常活动等特征。结果表明:提出的基于视频分析的人群密集场所风险预警技术框架可为城市大型商圈、交通枢纽、大型活动场所等城市公共场所的安全管理提供参考和借鉴。     

基于BP神经网络人群流量预测的实现

人群拥挤踩踏突发性的特点决定了现场的事故救援措施效果较差,事前预防是唯一有效的策略。对商业区人群流量进行预测,对于合理控制商业网点人口,预防人群类事故的发生具有重要的意义。本文介绍了基于BP神经网络的人群流量预测方法,利用Matlab建立了相关模型,并结合实际数据对模型进行了调整,分析了隐含层神经元个数、不同输入-输出结构、不同传递函数等因素对网络性能的影响。研究表明利用神经网络的非线性映射能力对人群流量进行预测时可行的。     

基于元胞自动机恐慌状态下人群疏散模型研究

为了研究恐慌情绪对疏散中人群行为的影响，结合决策理论和情绪感染相关理论，同时考虑恐慌情绪对行人决策的影响和恐慌情绪本身的动力学特性，如情绪感染和衰减等，采用元胞自动机方法，建立基于场域模型的恐慌状态下行人决策修正模型，采用该模型研究恐慌情绪对人群疏散行为的作用规律。研究发现:当处于优势疏散位置的行人，由于恐慌做出非理性决策，不仅降低了自身的疏散效率，而且会占据其他行人的最优疏散路径，因此导致整体疏散效率的下降；随着恐慌情绪阈值增加，同一时刻的恐慌人数减小，疏散时间减小，且存在某一临界值，使疏散时间最短；情绪自衰减速率是影响恐慌情绪的关键因素，且存在决定恐慌情绪的蔓延或者衰减的临界值；行人的恐慌程度随着行人密度的增加而增加，因为人群规模较大时，恐慌情绪的感染次数增加，并在人群中得到反复加强，从而使人均恐慌值增加。     

Study on mechanics of crowd jam based on the cusp-catastrophe model

Some characteristics of crowd jam like the phenomena of discontinuous jumping in reality are hard to be explained by the equations governing pedestrians. The catastrophe theory can explain these characteristics. A cusp-catastrophe model is developed to analyze the mechanics of crowd jam by drawing graphs for a cusp-catastrophe model of crowd, the bifurcation set and the projection of catastrophe model. Meanwhile, the critical density and the critical velocity are derived. It is concluded that the cusp model is a more efficient predictor than the linear model or the equations governing pedestrians and is reasonably realistic for dense crowd flow scenarios.     

Implementing real-time grouping for fast egress in emergency

Mass events such as the Olympic games or popular entertainment gathering often attract hundreds of thousands of people. If an abrupt event (e.g., fire, bomb threat) prompts people to rush to exits, stampedes may occur, creating secondary damages. Thus quickly dispersing crowd population to safe places in the least time while managing to avoid congestion and/or stampede is an important task with the occurrence of such events. This study presents an optimal evacuation model to meet this goal. By identifying some key themes that contribute to crowd disasters, the model implements a real-time concurrent grouping to ensure at-risk people are guided to safe places in the least time. The underlying rationale of the strategy is to advocate that competition should be replaced with coherence, and people should coordinate and collaborate to move out in an orderly way through scientifically using multiple routes. In doing so, each person receives an equal opportunity to exit the incident site, and the general effectiveness of performing an evacuation is achieved to an optimal level.To verify the proposed method, we conduct a simulation in which we compare our method with two benchmark cases: random self-evacuation and herding behavior. The results demonstrate that the proposed strategy can rescue people at risk much faster and more safely than two benchmark cases. We implement the proposed model through a web-based wireless decision support system, which provides a useful tool to both first responders and people affected for time-based crowd management in response to an ongoing crowd incident.     

基于多智能体建模方法的人群应急疏散模型构建研究——以腾龙芳烃（漳州）有限公司“4·6”爆炸着火重大事故疏散为例*

为从更微观角度分析人群疏散过程中疏散行为及路网设计对疏散效果的影响,基于腾龙芳烃（漳州）有限公司“4·6”爆炸着火重大事故,构建多智能体人群应急疏散模型,模拟人群中个体群组、惯性、就近、从众、信息传播的行为决策及相互交互影响。结果表明:群组行为会严重影响应急疏散效果,在疏散路网两端避难所附近会出现明显拥堵现象,在疏散路网设计和避难所选择时,应尽可能避免出现极端汇流路段或节点;在应急培训中,应告知群众减少群组行为。研究结果可为人群应急疏散提供借鉴。     

基于BP神经网络的河南省火灾风险评价

采用火灾发生起数、死亡人数、受伤人数、直接经济损失、烧毁面积、受灾户数及人口火灾发生率7项指标构建河南省火灾风险评价综合体系,并用BP神经网络对该省份的18个地级市行政区的火灾风险进行评价,评价结果显示,该方法不但可以体现各地区的火灾风险相对水平,而且还可避免传统方法的主观性,具有较强的可行性和可靠性,对控制火灾风险、客观地反映各地区的消防工作现状具有一定的指导意义。     

考虑人群拥堵的疏散出口选择行为研究及建模

出口选择是疏散过程中最重要和复杂的决策行为之一,受到空间结构、人群分布和行人认知等多方面因素的影响,为了使疏散仿真能够合理地模拟出行人的寻路和逃生过程,提出了考虑人群拥堵的出口选择模型。该模型基于多元Logit离散选择模型,考虑到影响个体疏散时间的因素以及对不同出口类型的偏好。研究发现个体的疏散时间随着拥堵人群位置的不同而有所差异,从而以决策者而不是出口的角度定义了一个计数区域来估计影响决策者排队时间的人数。并且研究出口重新选择的触发条件和程序。将出口选择模型和程序引入BuildingEXODUS软件进行疏散仿真,通过对比显示,仿真结果更符合真实情况,特别是在具有障碍物和人群分布不均匀的场景中更显优势。     

Computer and fluid modelling of evacuation

One of the most important areas in the field of “Life Safety” is the escape movement of individuals in emergency situations It is not yet possible to accurately model very complex psychological reactions, such as panic and confusion, but many parameters can be simulated, especially in the case of crowd movement. These basic parameters include speed fluctuations; crowd flow behaviour, travel distances and overall evacuation time, based upon certain assumptions. The evacuation characteristics of a building can be assessed in different ways. The statutory regulations (BSI, 1983) use general rules based upon the maximum crowd flow rates through specific exit route widths. These figures are based upon data produced in the 1950s (Hankin and Wright, 1958) that were intended primarily for application in building plans with fairly regular room/corridor plan configurations. This paper discusses two techniques for assessing crowd movement: fluid modelling and computer simulation. The fluid modelling is intended to provide a greater degree of understanding about the mechanism of crowd flow. The computer model SIMULEX combines spatial analysis with the escape movement of large numbers of individuals in a building, and is intended for eventual use as a design tool.