基于点采样的大规模三维模型显示

Point-Based Rendering of Large-Scale 3D Models

大规模三维模型的实时显示一直是许多虚拟现实应用追求的目标之一,也是计算机图形学领域的研究热点之一。文中采用基于采样点的显示方法,把具有显示质量优势的传统三角形与显示速度快、简化表示能力强的采样点,通过八叉树有机组织起来;通过对基本算法做出有效的改进,并添加计算分解和重用、有效的数据管理等模块,以高效的实现方式完成了30,000个独立运动的虚拟人场景的实时显示。文中的方法也可处理其它场景,能直接应用于紧急疏散或影视、游戏等领域中大规模三维模型的显示,具有很好的应用前景。

Real -time rendering of large -scale 3D models is the target for many virtual reality applications and meanwhile one of the hottest topics in the field of computer graphics. We adopt point - based policy for such a task in this paper, in which traditional triangles that produce good image quality are combined into an octree structure with samples which are good at representing simplified models with very fast rendering speed. We not only improve the original algorithm, but also add item decomposition and reusing, effective data management to it, and finally realize a real - time rendered scene of 30,000 individually animated virtual human models. The method is available for any kind of scene, and can be directly employed by emergency evacuation, video games, filmmaking et al. , which enjoys high application value with good prospect.