摘要
人群疏散仿真对于在紧急安全事故发生时指导人群快速、有序疏散,以及指导场景设计、提前制定应急预案等方面具有重要意义.但传统人群仿真方法在场景建模和场景语义提取的过程中存在复杂度高、低效的问题,影响了人群疏散仿真的效率.本文针对上述问题,提出一种基于地理信息的快速人群疏散仿真方法.首先,设计一种基于地理信息的场景建模方法,该方法能够从二维地图中获取道路的地理坐标信息,并利用几何变换、面片构建以及真实感处理技术快速得到场景模型;其次,定义基于地理信息的场景语义,并利用从二维地图中获取的地理信息建立路径拓扑图以提取场景语义,为人群运动计算提供路径规划和导航;最后,提出基于正态分布的相对速度障碍法(Normal Distribution-based Reciprocal Velocity Obstacles,ND-RVO),该方法在相对速度障碍法中加入正态分布速度,并结合场景语义和全局路径导航计算人群运动.实验结果表明,所提出的方法复杂度低,能够高效地进行人群仿真.
Crowd evacuation simulation has great significance, because it can guide crowd evacuation quickly and orderly in emergency situation. Meanwhile, crowd evacuation is also helpful for guiding architecture design, and developing emergency response plans in advance etc. But the traditional methods for crowd simulation are complex and inefficient both in scene modeling and crowd motion computing which affects the efficiency of crowd evocation simulation. To address the above problem,this paper presents a fast simulation method for crowd evacuation based on geographic information. First, we design a scene modeling method based on geographic information. The method can obtain the 3D scene modeling by abstracting the geographic coordinate of the road from 2D map and employing geometric transformation, surface construction and photo-realistic processing techniques. Then, we define the scene semantics based on geographic information and construct the topological graph to extract the scene semantics according to the obtained geographic information. The topological graph is used to path planning and individual navigation. Finally, we propose the Normal Distribution-based Reciprocal Velocity Obstacles method. The method combines normal distribution velocity with the Reciprocal Velocity Obstacles method and uses the scene semantics and global navigation method to compute crowd motion. Results show that this method has low complexity and can simulate crowd motion efficiently.
出处
《小型微型计算机系统》
CSCD
北大核心
2017年第10期2236-2241,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61572299
61402270
61272094
61472232
61373149
61402269)资助
山东省自然科学基金项目(ZR2014FQ009)资助
关键词
地理信息
快速场景建模
快速语义提取
人群运动仿真
geographic information
fast scene modeling
fast semantic extraction
crowd simulation
