【作者】 张春光；

【导师】 冯秀芳；

【作者基本信息】 太原理工大学 ， 计算机应用， 2004， 硕士

【摘要】 近几年来，随着各种计算机技术的飞速发展，虚拟现实得到了越来越多的应用。Internet网络走进千家万户，更是推动了分布式虚拟现实的发展。分布式虚拟场景的实时显示在虚拟现实中具有重要的地位，也是虚拟现实研究的重点和难点。而虚拟场景的建模和分块策略对场景的实时显示具有很大的影响，因此，虚拟场景的建模和分块策略成为虚拟现实研究的重要内容。本文重点对场景建模和场景分块的一些算法做了研究，并取得以下成果： 首先，对虚拟场景的各种主要建模方法做了研究。虚拟环境的建模是整个虚拟现实系统建立的基础，主要包括几何建模、运动建模、物理建模、对象特征建模等方法，每种方法各有优缺点。本文拟采用几种建模方法相结合的策略，对场景中逼真度要求比较高的对象，侧重于几何建模；场景中运动方式比较多的对象，侧重于运动建模等。这样产生的结果是：不要求在 太原理工大学硕士研究生学位论文某些方面特性突出，但可弥补采用单一建模方法产生的沉浸感弱等缺陷。 其次，对现有的场景分块策略作了详细的研究，现有的场景分块策略是建立在规整的正四边形基础之上的。因此，它的调度策略是当视点移动时，采用1+8+16策略。在蜂窝式电话服务网络的启发下，本文对现有的算法做了改进，提出了基于正六边形的场景分块策略，当视点移动时，与当前可见区相邻的6块最有可能成为下一个可见区，而外层的12个区是当前可见区的下一可见区的预可见区。所以在显示当前可见区时，保证对应的6个预可见区的场景数据信息己经处理完毕并与当前可见区的场景一起存放在内存中。而这同时并行处理12块将来预可见区的数据。本文也就称这种策略为卜6+l2策略，即:显示1块，调入6块，计算12块。优点是:场景显示时，空间复杂度明显降低了。提高了场景显示的实时性和沉浸感。 再次，场景采用一维数组链表存储。本文提出的X+Y存储算法，是对现有二维数组存储方式的一种改进。相当于在原来的基础上加了一级索引。提高了场景块的查找速度，提高了定位的速率。增强了实时性。太原理工大学硕士研究生学位论文最后，本文对单块场景的建模也提出了一点改进。更多还原

【Abstract】 Recently, with the rapid development of the computer technology, virtual reality has been used in more and more fields, and with the popularization of Internet, the evolution of the distributed virtual reality has been accelerated. As the emphases and the difficulty of the research about the virtual reality, to display the scene of the distributed virtual environment in time plays an important role, and the modeling and partition strategy of the virtual scene can affect the real-time display of the scene largely, so they are main content of the research of the virtual reality, we stress some arithmetics about the modeling and partition strategy, and we get some achievements as follows:Firstly, the research of main modeling method in virtual scene has been given. Usually, the existing modeling method is based onthe use of the single or some of the modeling methods, such as geometrical modeling, motorial modeling, physical modeling and object character modeling. So the result is: the characteristic of one side for the scene is prominent, but the other side is very poor. In this thesis, we use all these modeling methods, geometrical modeling to objects which need high realistic, motorial modeling to which require more movements. Then the result will be better: no prominent on one side, but the limitation of the method we mentioned is improved.Secondly, we research the partition strategy to the existing scene in detail which is based on the regular quadrangle. So its attempering strategy is to adopt "1+8+16"strategy when the viewpoint moves. Under the elicitation of honeycomb telephone service net, we ameliorate the existing arithmetic, propose a partition strategy based on regular hexagon. When viewpoint moves, there are six blocks near to the current visible area may be the next one, while the outer 12 blocks may be the pre-visible area of the next visible area. So while displaying the current visible area,we should make sure that the scene data information of the corresponding 6 pre-visible areas has been operated, and they have been saved into memory together with that of the current visible area. Meanwhile, the data of 12 pre-visible areas been operated parallelly. This strategy can be called "1+6+12" strategy in this thesis. That is: to display 1 block , fold 6 blocks, compute 12 blocks. Its merit is the special complexity has been reduced evidently when the scene displayed , and the real-time ability and immersion has been improved.Thirdly , the scene is saved in the array linking list in one-dimension, we bring forward X+Y saving-arithmetic which is an improvement for the existing array in two_dimension. That is to say we add the first-refrence on the base of the former , which quickens up the lookup for the scene block , improves the speed of the orientation and boosts up the real-time ability.Lastly, we put forward a little improving advice for modeling to the single block sence in this thesis.更多还原