Large data sets of particle data even today pose a big challenge for visual analysis methods. Explorative analysis is important to discover unexpected effects inside the data. For this kind of analysis, interactive visualisation provides is the optimal method. Subproject D.3 of SFB 716 faces the challenge of visualising large, time-dependent particle data sets and develops new methods and algorithms to reach this goal. Major enhancements to particle visualisation are needed as well as close cooperation with domain experts inside the SFB to accomplish this task.
Massive effort has been put into the development of particle-based visualisation in multiple SFB subprojects. The point-based visualisation approach as well as GPU-raycasting of implicit surfaces makes it possible to interactively visualise data sets containing millions of particles at high quality. The large size and the time-dependency of these datasets as well as the data transfer between secondary storage, main memory and graphics memory become a severe bottlenecks, wherefore transfer methods have been studied and optimised. A two-phase culling approach has been developed and optimised for particle-based data which allows for interactive representation of 107 to 109 particles on a standard workstation.
To improve the quality of the visual representations beside rendering speed, deferred shading technqiues have been extended. The aspect of optimal rendering methods for complex glyphs, like dipole glyphs, composite glyphs for molecules without inner degrees of freedom and polyhedron-based glyphs for porous media has been investigated as well. To enhance perception and understanding of the presented data, filtering is needed, like for the visualisation of crystal defects in metals. The formation fo certain structures inside data sets are of special interest if we want to observe temporal evolution of this data. To compensate for overlap and visual overload when using path lines the data set has been filtered using new density-conserving clustering techniques. Based on this work other static representations of dynamic phenomena will be developed for other subprojects inside SFB, allowing more effective analysis.
To this end, an extensible and adoptable visualisation framework has been developed during the first funding period of SFB 716: MegaMol™. Highly optimised renderers and data structures form the basis for the current visualisation research for this subproject, as well as for the visualisation subprojects D.4 and D.5 of the SFB. Thanks to the modular architecture of the software, adaptable at run time, and the extensibility through a plug-in interface and simple programming interfaces, MegaMol™ is optimally suited for the wide range of different demands of the subprojects of SFB 716.
MegaMol™ succeeds MolCloud, which has been developed at the University of Stuttgart in order to visualise point-based data sets. MegaMol™ is written in C++ and uses an OpenGL as rendering API as well as GLSL shaders. It runs on Microsoft Windows (Vista, 7 and 10) and Linux (Suse), each in 32-bit and 64-bit versions.
- Collaborative Research Centre (SFB) 716 – Dynamic simulation of systems with large particle numbers
Project D.3 - Visualisation of systems with large numbers of particles
Long-term goal of project D.3 is the development of methods for interactive, visual analysis and exploration of simulation data sets with large numbers of particles and long trajectories.
MegaMol is a visualisation middleware for point-based molecular data sets.
- DLR Landesstiftungsprojekt 688: Visualisierung der Keimbildung in Mischungen für skalenübergreifende Modelle (completed)
This project formed the foundations for projects D.3 and D.4 of SFB 716 by providing essential work in the field of interactive visualisation of molecular dynamics simulation data. The funding by Landesstiftung Baden-Württemberg ended at the end of 2006.
- MolCloud (completed)
MolCloud is a visualisation utility for molecular dynamic data sets, developed at the University of Stuttgart. In fall 2006, the project was ended and the work its successor on MegaMol™ began.
- PointCloud (completed)
PointCloud accelerates rendering of scattered point data by means of a a hierarchical data structure based on a PCA clustering procedure and thus formed the starting project of point-based visualisation at our institute.