面向场景模型干涉的实心体素化方法  

Solid Voxelization Method for Scenes with Interfered Models

作　　者：路强[1] 杨柳青[1] 李琳[1] 李尚林[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《系统仿真学报》2014年第12期2921-2927,共7页Journal of System Simulation

基　　金：国家自然科学基金项目(61070124);高等学校博士学科点专项科研基金(20120111110003)

摘　　要：场景表示是虚拟现实系统中的核心内容之一,体素是一种有效的场景表示方法。很少有研究关注场景中存在模型干涉的情况,如模型相交、模型裁剪,此类场景难以正确填充。提出了一种近似生成均匀体素体的并行GPU算法,使用逐体素列有序链表来存储光栅化产生的几何模型片段,通过对链表中几何片段进行配对来确定模型边界,以达到可以正确区分体素是在模型内部还是外部的效果。实验证明该方法可以正确处理存在模型干涉的模型内部区域,可满足实时程序的性能要求。Scene representation method is one of the core parts of virtual reality systems and the voxel representation is an effective one among them. Scenes with interfered models, intersected and clipped for example, are common in a real scenario, but few researches focus on this. As a result, no effective voxelization methods exist for these scenes. An efficient parallel GPU algorithm was proposed to approximately generate uniform solid voxel volume for real-time applications. The method used per voxel volume column linked lists to store rasterizer generated geometry fragments, then a pairing process was used with these fragment to resolve whether a voxel is inside/outside a model. As the experiment results show, the method can fulfill the interior region of models with interfering correctly, and it can satisfy the performance requirements of real-time usage.

关 键 词：实心体素化 实时 GPU 光栅化 逐体素列链表 片段配对 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]