Bag-of-particles as a deformable model.

摘要

The computer graphics community has historically focused on three aspects of model realism (visual, behavioral, interaction) individually and to varying degrees, with extensive research into the former, but scant research into the latter. Over the last two decades, increasing attention has been directed to physically-based deformable models, with several different fundamental approaches emerging, each with strengths and weaknesses. In particular, physically-based models have used particles in a number of specialized ways. Particles have been employed to create deformable surface shapes that are devoid of interior material, which is ignored by the model. Systems modeling amorphous, bulk matter have been developed, but in these systems various properties of the particle ensemble are of interest, and surface shape is ignored. In some systems particles are fixed to polygon vertices, requiring that a polygonal object representation exist. In general, particle systems have dealt with homogenous material only, and they have not addressed the issue of haptic modeling.; This thesis presents a physically-based, elastically deformable model using a particle system to model surfaces with interior volumes that can be haptically felt. Oriented particles used in existing surface-only models, and unoriented particles used in volume-only simulations are combined to form a bag-of-particles. Multiple particle species and predefined interspecies parameters determine the elastic properties of a bag, permitting objects composed of non-homogenous material, such as organs with multiple tissue types, to be modeled. Bag-of-particle models are constructed from a 3D voxel bitmap object represention, which can be derived from various types of source data. This single representation as a bitmap is used to capture the static shape of an object, and to derive a distance map which provides both shape-maintaining forces and feedback reaction forces that allow a user to “feel” the model through direct manipulation. By combining a distance map to represents shape, and a particle system to model physical behavior, the behavioral and interaction models are integrated in a single framework. The approach is demonstrated using several different initial object representations, and is applied to a dynamic model of a human heart ventricle through one complete cardiac cycle.