中文摘要：

大规模群体的三维可视化是虚拟现实领域的研究热点之一．目前，由于驱动方法与渲染效率的问题，在虚拟空间中创建的大规模虚拟群体很难同时满足驱动与渲染过程的逼真与实时．该问题对于群体规模庞大，个体外形、动作具有一定程度个性化，需要进行独立运动控制的场景尤为明显．针对这一问题，提出了一个高效的大规模虚拟群体三维可视化方法：首先，通过使用模板派生技术从少量模板模型派生出大量外观各异的个体模型；其次，使用运动数据个性化变形技术实现了运动数据的重用；最后，使用基于采样点的渲染技术实现了大规模数据的实时显示．在上述工作基础上，研究并开发了一套大规模虚拟群体的三维可视化系统，能够方便的在普通PC机上实时、逼真的展示大规模的运动虚拟群体，并实现了60000群众紧急疏散过程的实时三维可视化．

英文摘要：

3D visualization of large-scale virtual crowds is a very important and interesting problem in research fields of virtual reality. For reasons of efficiency and visual realism, it is very difficult to populate, animate and render large-scale virtual crowds with hundreds of thousand individually animated virtual characters in real-time applications; especially for the scene which has more than ten thousands of individuals with different shapes and motions. In this paper, an efficient method to visualize large-scale virtual crowds is presented. Firstly, using model variation technique, many different models can be derived from a small number of model templates. Secondly, the crowds can be animated individually by deforming a small number of elements in motion database. Thirdly, using a developed point sample rendering algorithm, large-scale crowds can be displayed in real-time. This method can be used to visualize different dynamic crowds which require both real-time efficiency and large number of virtual individuals support. Based on this work, an efficient and readily usable 3D visualization system is presented. It can provide very high visual realism for large crowds＇ visualization in interactive frame rates on a regular PC. The 3D visualization of 60000 people evacuating from a building is also realized in real-time based on the system.