An exact general remeshing scheme applied to physically conservative voxelization

TL;DR

This paper introduces an exact remeshing scheme that conserves physical quantities like mass and energy during voxelization, using robust geometric algorithms for intersecting convex polyhedra, applicable in physics, graphics, and mechanics.

Contribution

The authors develop a robust, exact remeshing algorithm for convex polyhedra that guarantees conservation of integrals, improving accuracy and stability over previous methods.

Findings

Ensures mass, momentum, and energy conservation during remeshing.

Achieves substantial performance improvements over recent methods.

Provides a robust C implementation for geometric calculations.

Abstract

We present an exact general remeshing scheme to compute analytic integrals of polynomial functions over the intersections between convex polyhedral cells of old and new meshes. In physics applications this allows one to ensure global mass, momentum, and energy conservation while applying higher-order polynomial interpolation. We elaborate on applications of our algorithm arising in the analysis of cosmological N-body data, computer graphics, and continuum mechanics problems. We focus on the particular case of remeshing tetrahedral cells onto a Cartesian grid such that the volume integral of the polynomial density function given on the input mesh is guaranteed to equal the corresponding integral over the output mesh. We refer to this as "physically conservative voxelization". At the core of our method is an algorithm for intersecting two convex polyhedra by successively clipping one against the faces of the other. This algorithm is an implementation of the ideas presented abstractly by Sugihara (1994), who suggests using the planar graph representations of convex polyhedra to ensure topological consistency of the output. This makes our implementation robust to geometric degeneracy in the input. We employ a simplicial decomposition to calculate moment integrals up to quadratic order over the resulting intersection domain. We also address practical issues arising in a software implementation, including numerical stability in geometric calculations, management of cancellation errors, and extension to two dimensions. In a comparison to recent work, we show substantial performance gains. We provide a C implementation intended to be a fast, accurate, and robust tool for geometric calculations on polyhedral mesh elements.