随着城市人口不断增加，交通网络的密集，多功能的综合体建筑的迅猛发展，城市中公共场所的行人变得越发密集，再加上大型活动的数量、规模逐渐增大，使得密集人群的安全管理受到了高度的重视。因此，本文分析了行人在不同密度环境下的运动规律和行为模式，研究了人群聚集风险评估方法，从而为公共场所的设计、监测监控及人群中踩踏事故的预测预警提供帮助，为科学有效的应急预案的编制提供指导意见。 行人在不同密度下的行为特征显著不同。本文首先基于单列行人运动实验着眼于低密度时行人的运动特征和行为模式，提出了基于行人运动轨迹的迈步特征的识别、计算方法，得出了迈步特征与速度的定量关系，发现了身高对迈步特征的影响会随着密度而改变。另一方面，发现了行人的两种自适应行为，一是通过改变迈步方式适应周围环境，行人除了常见的等步长迈步方式外，在高密度情况下会出现大小步交替进行的迈步方式；二是行人之间保持同步，发现了两种同相同步现象，研究了同相同步现象对行人步长的影响。 密集人群中除了独自行走的人群外，还存在成组运动的行人，本文基于实际案例的视频数据、空间位置数据及群组信息数据，挖掘了群组在单向、双向运动环境下同组行人之间的空间关系，研究了群组在不同密度环境和空间关系下所需要的行走空间，分析了群组规模对行人之间距离的影响，发现了群组的两种自适应行为：通过改变相互之间的距离和相对位置来适应周围的环境，群组所需要的空间与行人的密度和群组的空间关系有关，群组成员之间以及与组外行人之间的距离都与群组规模成正比例关系，且密度越大，相应的增长率越低。 上述工作是关于低密度下行人运动特征的研究，而密集人群的运动特征及其研究方法都与之不同，因此针对高密度人群，本文基于6个实际案例的视频资料，包括无突发事件触发的有序、无序人群和有突发事件触发的骚乱人群，对比分析了高密度人群有序运动、无序运动的速度、加速度特征，发现不同场景下湍流人群的运动特征不同，确定了将加速度作为表征人群聚集风险的参数，从时间、信息、空间3个维度研究了风险分级方法，提出了适应于无外界突发事件触发和有外界突发事件触发的人群聚集风险评估方法。

With the growth of urban population as well as the development of the transportation and multi-functional urban complex, the public places become more and more crowded. Besides, large-scale mass activities were organized much larger and more frequently. In these situations, the pedestrian safety attracts much attention. Thus to predict pedestrian behavior and give early alarm for stampede, the movement characteristics of pedestrians at low and high densities were analyzed in this paper. The risk assessment method for crowd gathering was studied based on real mass activities. The corresponding results can help design pedestrian facilities, improve the monitoring technology of crowded places and make effective emergency plans. Pedestrians behave differently at various densities. Firstly experimental study on the single-file movement of pedestrians at different densities was conducted. The gait characteristics, speed and density were calculated based on the trajectories of pedestrians. The dependences of gait characteristics on speed, density and body height were investigated. It is shown that the effect of body height on step length and step duration changes with density. Furthermore, two self-organized movement patterns were found. One is that pedestrians adjust their step styles to the limited space. Six different step styles were observed in this experiments. The other is the step in-phase synchronization between two successive pedestrians. Besides pedestrians walking alone, some people walk in groups. To study pedestrian grouping behaviors, the groups in the pedestrian flow were recognized by observing the video recordings of two empirical cases. Their trajectories were obtained via PeTrack and 3-D range sensor respectively. The spatial patterns and required space of groups at different densities were investigated. It is found that groups change their spatial patterns and the distance between each other to avoid collision or adapt to the environment. The required space of groups is influenced not only by their density but also by the spatial patterns. The distance among group members and the distance between the group member and the pedestrian outside the group both increased with group size linearly. The growth rate decrease with the rise in density. The studies mentioned above are about pedestrian movement behaviors at low densities. However, the crowd at high densities show collective behaviors and the corresponding analysis methods are different. To investigate the characteristics of highly densed pedestrians, the video recordings of six different real mass activities were collected and analyzed via the speed extraction algorithm based on cross-correlation algorithm. The spatial and temporal distributions of velocity and acceleration characteristics were studied. It is found that the turbulent movements vary under different situations, which are influenced by the propagation direction of shock wave and facility type. The accelerations of pedestrians can detect the risk in the crowd. A new risk evaluation approach for crowd gathering is proposed, which can apply to the situations with and without emergencies.