【作者】 聂永丹；

【导师】 刘贤梅；

【作者基本信息】 大庆石油学院 ， 计算机应用技术， 2009， 硕士

【摘要】 LOD技术在虚拟现实中的应用越来越广泛,通过对虚拟环境中非重要场景的多细节层次表示,减少实时渲染的计算花费,目前在几何层的发展已经达到成熟阶段,运动层、行为层的LOD技术研究已经成为虚拟现实技术的一个新的研究方向,其研究的重点是对虚拟环境中的运动角色进行多细节层次表示。本文介绍了目前虚拟环境中几何层典型的LOD简化技术;在运动层,对运动角色的表示方法进行了分析,对运动层应用LOD思想进行简化所需要的运动编辑技术、骨骼简化技术、关节点姿势聚类技术进行了研究,并提出在运动层提取运动特征的方法对运动捕获数据进行标注,实时渲染时根据运动捕获数据的标注采用不同的简化方法,将运动层的部分简化工作放在了预处理阶段,提高了实时处理速度;在行为层,提出通过为角色建立基本行为模型和复杂行为模型实现行为层LOD思想的应用;同时本文提出了虚拟环境中基于几何层、运动层、行为层的LOD简化框架,便于用户通过该框架建立虚拟角色行为模型的同时实现了虚拟环境中三个层次的LOD简化。本文采用运动捕获数据驱动骨骼运动,通过对BVH文件的扩展对运动层离散LOD简化进行了尝试,同时在运动层提出了基于帧差距控制连续LOD简化的方法,并将其应用到舞蹈方阵中,取得了良好的视觉效果,同时降低了渲染时间。实验表明,通过LOD技术对虚拟场景中的运动角色进行多细节层次表示,能够在保留角色运动主要特征的基础上,提高渲染速度,在帧频率恒定的情况下,该方法是提高虚拟环境中角色数目的有效方法。更多还原

【Abstract】 The application of LOD technology in Virtual Reality is more and more widely, by which represents non-important and important geometric objects in virtual environment with different level of details to reduce the computation cost of real-time rendering. The geometric LOD algorithms have already been researched for many years , then the research on motion LOD and behavior LOD becomes a new research direction of the virtual reality technology, and it focus on representing the animation characters with different level of details of motion and behavior.This paper introduces the LOD simplification techniques at geometric level of the virtual environment and analyzes the represents of animation characters at motion level. The technology of LOD simplification at motion level, such as motion edit, skeleton simplify, Joint Posture Clustering, is discussed, We present the methods extracting motion characteristics to mark motion capture data at motion level, and select different simplify methods based on these marks when rendering. By this way, many simplifying works at motion level are put in the pre-process stage. The described approach accelerates the rendering speed in real-time. At behavior level, the basic behavior models and complex behavior models are proposed; A LOD simplify framework in virtual environment based on the geometric, motion, behavior is presented, by which can create virtual character behavior models and implement the LOD simplify at three levels.This paper uses motion capture data to driven skeleton animation. The discrete motion LOD simplify method is experimented by extending BVH file, then the continuous motion LOD simplify method is controlled by frame distance. Above methods are applied to dance square application, and achieves better visual effects while reducing rendering time.The experimental results show that it can retain main features of character animation and improve rending speed by LOD technology that represent animation characters with different level of details of geometry, motion and behavior. The proposed method is effective to increase the number of characters in virtual environment at a constant frame rate.更多还原