Physics-based subspace deformation: Theory and application.

详细信息   

• 作者：Yang ; Yin.
• 学历：Ph.D.
• 年：2013
• 导师：Guo, Xiaohu,eadvisorMao, Weihua,eadvisorDaescu, Ovidiuecommittee memberMcMahan, Ryan P.ecommittee memberPrabhakaran, B.ecommittee memberZhang, Kangecommittee member
• 毕业院校：The University of Texas
• Department：Computer Science
• ISBN：9781303332159
• CBH：3592233
• Country：USA
• 语种：English
• FileSize：14769994
• Pages：114

文摘

Fast physics-based modeling and simulating is important for realism, which makes games more interactive and movies more vivid. Compared with earlier graphics applications, physics is more and more involved and demanding in state-of-the-art applications. Elastic deformation is one of the most commonly-seen yet challenging topics among the various types of physics. The most physics-rigorous approach to model 3D elastic deformation is taking use of strain-stress relationship e.g., Hookes law) based on the stress equilibrium analysis of the soft object. The equilibrium is typically described with a set of partial differential equations and the whole 3D domain of the soft object is further discretized with mutual connected finite elements which are referred as FE mesh. The domains deformation is then expressed with the displacement at each FE node on the mesh. Finite element method FEM) is quite accurate compared with other methods i.e., mass-spring model). Meanwhile, its shortcoming, that FEM generally needs more computation is also obvious and the high computational demand makes online-computed deformation rarely seen for real-time applications. Spectral model reduction is such a technique that uses a small number of basis deformation shapes to describe the deformed soft object. We also present a novel algorithm to further boost the pre-computation with a divide-conquer approach to handle by dividing the complex mesh into multiple sub-domains. An instant benefit is that the domain-based pre-computation becomes remarkably faster than the mesh-based one. A nice byproduct of this approach is that the simulation itself is even faster than the single domain spectral model reduction. Medical and biomechanical modeling involves a lot of deformation analysis as most human organs are soft deformable objects. The elastic deformable model often captures the natural features of the organs better than pure geometry or signal based approaches. Modeling heart beating, tumor growth, muscle tension, and numerous other medical and biomechanical research problems require such combination of kinematics, dynamics, mechanics, image analysis and graphics. Undoubtedly, the framework of fast deformation simulation provides good facilities to these problems.