MPI-INF Logo
Campus Event Calendar

Event Entry

What and Who

Validation of a State-of-the-Art Fluid Simulation with a 3-D Animation Scanner

Philipp Jenke
Max-Planck-Institut für Informatik - AG 4
AG4 Group Meeting
AG 4  
AG Audience
English

Date, Time and Location

Tuesday, 15 February 2005
13:00
45 Minutes
46.1 - MPII
019
Saarbrücken

Abstract

In recent years, modern fluid simulations have been an important research topic in the field of computer graphics. Several new techniques have been developed and their results look more realistic than ever before. However, computer graphics' applications put the major emphasis on the quality of the resulting images rather than on the validation of the physical correctness. Therefore, it seems natural to evaluate these fluid simulation methods in terms of physical reality. As the scanning technology has improved significantly, researchers for the first time have the required tools to compare fluid simulations with real phenomena. Consequently, the comparison of a scanned fluid with a state-of-the-art fluid simulation is presented in this work.


A fluid of medium viscosity is captured using an optical animation scanner. This scanner acquires the surface of an arbitrary object from a fixed camera perspective. The scanning technique allows to capture real fluid behavior as its frame rate is high enough to extract a movie. After being separated from the rest of the scanned data the fluid's surface is converted into a triangle mesh.

The simulation is based on one major differential equation, namely the Navier-Stokes equation. Velocity values are defined all within a staggered regular grid. They are integrated forwards in time using a semi-Lagrangian approach and finite differences. The actual fluid is then represented via a particle level set method using a discretized iso function. This representation allows to move the surface in the velocity field accurately. A triangle mesh of the surface is extracted using the Marching Cubes algorithm.

For the purpose of comparison, different approaches are discussed. A height field representation is chosen then, as height fields are easy to extract from both datasets. Additionally, the resulting height fields are very suitable for efficient comparison. The results show that it is possible to clone the physical reality with the tools of a fluid simulation. Some future work will have to be appointed to the comparison of other fluid features like internal velocities or the viscosity.

Contact

Volker Blanz
--email hidden
passcode not visible
logged in users only

Volker Blanz, 02/02/2005 14:09
Volker Blanz, 02/01/2005 18:10 -- Created document.