Automatic Conversion of Animated Models to Volumetric Representations with Runtime Geometry Modification

Date of Award

5-2026

Degree Name

Master of Science

Department

Computer Science

First Advisor

Li Yang, Ph.D.

Second Advisor

Steven Carr, Ph.D.

Third Advisor

Nada Lachtar, Ph.D.

Keywords

Computer graphics, destructable, real-time rendering, skeletal animation, volumetric

Access Setting

Masters Thesis-Abstract Only

Restricted to Campus until

5-1-2036

Abstract

Current approaches for rendering models in computer games represent the model as a polygonal surface mesh. This limits their ability to render penetration effects such as wounds, to model accurate ballistic behavior, and to render simulations of destruction effects such as breaking walls or cutting into wood. Automatic volumetric conversion has traditionally been infeasible due to existing tetrahedralization algorithms requiring manifold inputs free of self-intersections. Most models also require animation support which depends on skinning weights. An approach to adapting skeletal rigging data from an input model to a volumetric representation is introduced as well as a technique for automatic conversion of texture coordinates. An efficient GPU-oriented solution is developed to enable dynamic modification of the geometry of models at runtime for use in rendering effects like penetration. These contributions in conjunction with recent advances in tetrahedralization and geometric data interpolation are used to introduce a pipeline for automatically converting real-world models to volumetric representations. Results demonstrate that the implementation handles real-world models with defective geometry while preserving animation fidelity.

This document is currently not available here.

Share

COinS