数据驱动的人群场景动画合成方法  

A Data-Driven Method for Crowd Simulation in Urban Scenes

作　　者：王鑫[1] 孙守迁[1] 侯易[1] 邓春燕[1] 

机构地区：[1]浙江大学计算机科学与技术学院,杭州310027

出　　处：《计算机辅助设计与图形学学报》2009年第2期250-256,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金重点项目(60533080);浙江省科技计划项目(2007C23096);杭州市产学研项目(20062612E04)

摘　　要：为了向虚拟城区环境加入复杂而灵活的人群动画,提出一种数据驱动的人群场景动画合成方法.通过采集各种典型场景下人群流动的视频,提取出agent状态响应数据集;并将这些状态响应数据集嵌入到虚拟城区知识模型中.最后基于该虚拟城区知识模型实现了一个虚拟城区中人群流动场景的动画合成系统.This paper proposes a new data driven method for crowd simulation, which can be used for adding scalable and complicated crowd animation into a virtual urban scene. By the method firstly,we capture several typical pedestrian movement videos, from which state-action data sets are extracted. The data sets are then integrated into an informed virtual urban model. And finally, a crowd animation synthesis system of pedestrian in the virtual urban is established based on the informed virtual urban model.

关 键 词：群体仿真 行为动画 城区仿真 视频跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]