基于数字图像处理的人员运动参数提取方法研究  

Method for extraction of dynamic parameters of pedestrian traffic based on digital image processing

作　　者：刘轩[1] 宋卫国[1] 马剑[1] 张俊[1] 

机构地区：[1]中国科学技术大学火灾科学国家重点实验室,安徽合肥230026

出　　处：《火灾科学》2008年第4期201-208,共8页Fire Safety Science

基　　金：国家自然科学基金(50678164);国家科技支撑项目(2006BAK06B00)资助

摘　　要：针对人员疏散研究中提取疏散场景行人运动参数的需要,采用数字图像处理技术对记录行人运动的视频进行了处理。首先对背景进行混合高斯建模并提取前景,然后采用mean-shift算法对视频中彼此相连的多个行人目标进行有效地聚类分割并准确地跟踪每个行人,得到了包含有行人运动特性时空信息的运动轨迹,利用轨迹得到了行人在行进过程中的速度变化特征。结果表明该文所提出方法可以方便、快速、准确地提取人员运动特性参数,可望对提取人员疏散实验以及真实场景中行人运动的基础数据,如速度、密度、流量等提供有效的技术手段。In order to extract dynamic parameters for evacuation research, in this paper, we process the video scene of pedestrian movement. Firstly, we extract the foreground using mixed Gaussian models. Secondly, we use mean--shift algorithm to effectively segment and exactly track the connected pedestrian objects in the video which records the people moving with a free velocity. As a result, we get the trajectories which contain all the spatial and temporal information about the pedestrian movement in the video scene. And through the trajectories of pedestrians, we obtain the velocity variation characteristics of pedestrians who move with a free velocity. The results show that the method can extract characteristic parameters of pedestrian movement easily, rapidly and exactly. This method can provide effective means to extract fundamental data such as the velocity, pedestrian density and flow in the evacuation experiments and real evacuation scenes.

关 键 词：人员疏散 运动参数 混合高斯模型 mean—shift算法 

分 类 号：TU972[建筑科学—建筑设计及理论]