Collision detection and post-processing for physical cloth simulation

DSpace Repositorium (Manakin basiert)


Dateien:

Zitierfähiger Link (URI): http://nbn-resolving.de/urn:nbn:de:bsz:21-opus-17890
http://hdl.handle.net/10900/48769
Dokumentart: Dissertation
Erscheinungsdatum: 2005
Sprache: Deutsch
Fakultät: 7 Mathematisch-Naturwissenschaftliche Fakultät
Fachbereich: Sonstige - Informations- und Kognitionswissenschaften
Gutachter: Straßer, Wolfgang
Tag der mündl. Prüfung: 2005-06-29
DDC-Klassifikation: 004 - Informatik
Schlagworte: Computersimulation , Kollisionserkennung , Computergraphik
Freie Schlagwörter: Textilsimulation , Nachbearbeitung
Simulation , Cloth , Collision Detection , Post-Processing
Lizenz: http://tobias-lib.uni-tuebingen.de/doku/lic_mit_pod.php?la=de http://tobias-lib.uni-tuebingen.de/doku/lic_mit_pod.php?la=en
Gedruckte Kopie bestellen: Print-on-Demand
Zur Langanzeige

Inhaltszusammenfassung:

Um deformierbare Materialien wie Textilien oder menschliches Gewebe physikalisch simulieren zu können, muss eine Reihe komplexer Probleme gelöst werden. Diese Arbeit befasst sich dabei im Speziellen mit der Detektion und Behandlung von auftretenden Kollisionen sowie der Nachbearbeitung simulierter Netze zur Steigerung der visuellen Qualität. Dabei werden Anwendungen aus dem Bereich simulierter Textilien und der virtuellen Anprobe von Kleidung vorgestellt. Insbesondere wird auf das am WSI/GRIS entwickelte Textilsimulationssystem TüTex eingegangen. Zunächst wird der State of the Art zur Kollisionsdetektion deformierbarer Objekte vorgestellt, und es wird ausführlich diskutiert, welches Verfahren am besten für welche Anwendung geeignet ist. Für die Anforderungen, die TüTex an die Kollisionsdetektion stellt, wird gezeigt, dass sich dafür besonders Bounding-Volume-Hierarchien eignen. Diese werden im Weiteren mit einem stochastischen Sampling zu einer neuen Kollisionsdetektionsmethode kombiniert. Dieses neue Verfahren erlaubt eine Abwägung zwischen Geschwindigkeit und Qualität der Detektion und erhöht damit auch deutlich ihre Performance. Im Folgenden wird eine Impuls-basierte Methode zur Auflösung komplexer Kollisionen und Selbstkollisionen vorgestellt, die sowohl für statische als auch dynamische Kollisionsobjekte stabile Simulationen sicherstellt. Da Textilsimulationen mit hoch aufgelösten Netzen nach wie vor sehr zeitintensiv sind, wird vorgeschlagen, grobe Netze zu simulieren und diese anschließend geometrisch nach zu bearbeiten. Dazu trägt diese Arbeit zwei Verfahren bei. Um die sichtbare polygonale Struktur von groben Netzen zu beseitigen werden Subdivisionmethoden benutzt. Dabei werden interpolierende und approximierende Verfahren in Bezug auf die Eignung bei virtuellen Textilien verglichen. Da der Subdivisionschritt selbst auch wieder zu Kollisionen vor allem zwischen dem modifizierten Textilnetz und seiner Umgebung führen kann, werden diese Verfahren mit einer kontinuierlichen Kollisionsdetektion und Kollisionsantwort kombiniert. Als zweites Verfahren zur Nachbehandlung virtueller Textilien werden Faltentexturen vorgeschlagen. Da grobe Netze keine feinen Falten modellieren können, werden diese Details durch Texturen hinzugefügt. Diese Texturen werden dabei auf Basis der Deformation der Netze generiert und können als Bump- oder Displacement-Map verwendet werden. Im Gegensatz zu früheren Verfahren wird diese Faltentextur ohne Benutzerinteraktion generiert, was die Verwendung in automatischen Simulationssystemen wie TüTex ermöglicht. Des Weiteren werden in dieser Arbeit erstmals diese Texturen mit der kontinuierlichen Kollisionsdetektion kombiniert, um ein kollisionsfreies Displacement-Mapping zu realisieren. Beide Nachbearbeitungsverfahren führen zusammen mit groben Netzen zu visuell vergleichbaren Ergebnissen wie die Simulation hoch aufgelöster Netze ohne Nachbearbeitung, ermöglichen aber deutlich kürzere Simulationszeiten. Die in dieser Arbeit entwickelten Konzepte wurden in mehrere Systeme zur Kleidersimulation integriert. Mit Virtual Try-On wurde dabei das erste System umgesetzt, welches die physikalisch-basierte Simulation von Kleidung basierend auf CAD-Schnittmustern, physikalischen Materialparametern und 3D-Körperscans ermöglicht. Außerdem wurde der Textilsimulator TüTex als Plugin für die Modellierungssoftware Alias Maya weiter entwickelt, um über eine effiziente und komfortable Test- und Visualisierungsumgebung zu verfügen.

Abstract:

The objective of this thesis is to develop collision detection, collision response and post-processing algorithms for the cloth simulation system TüTex. In the following we first summarize our specific contributions to these fields, before we address possible research directions for the future. Collision Detection and Response: For collision detection methods employed for deformable objects we presented the state of the art and extensively discussed the respective advantages and disadvantages. We gave a decision matrix that allows to choose the method that fits best to the specific needs of a problem or an application. Using this decision matrix, we motivated the choice of bounding volume hierarchies for the cloth simulation system TüTex. As bounding volume hierarchies per se do not allow a balancing of speed and quality during the collision detection process, we developed an innovative stochastic collision detection method. This hierarchy accelerated stochastic collision detection shows a significantly better performance than a pure stochastic or a pure bounding volume hierarchy approach. To motivate the balancing of speed and quality, we showed that it is not necessary to detect all collisions in order to obtain stable cloth simulations. To resolve emerged collisions we developed an impulse-based response method that is able to handle complex collision and self-collision situations both for static and for dynamic collision objects. Additionally, we showed how the applied impulses of this method are distributed on the single mesh vertices in order to obtain continuity across mesh elements. Post-Processing: Since cloth simulation with high resolution meshes is very time consuming, we proposed the simulation of low resolution meshes combined with a subsequent post-procession. To the field of post-processing of virtual textiles we contributed two different approaches. To overcome the problem of the visible polygonal structure of coarse meshes, we proposed to use subdivision methods for triangular meshes and compared the approximating modified Loop and the interpolating modified Butterfly method with respect to their applicability for virtual cloth. As the refinement of the mesh using subdivision methods may cause collisions between the altered cloth mesh and its environment, we combined the subdivision methods with a continuous collision detection and response. The second post-processing approach we introduced in this thesis are the strain-based wrinkle textures. Since coarse meshes cannot model fine folds and wrinkles, these details are added by textures. Based on the assumption of area conservation within a triangle and the strain tensor as a deformation measure, a texture representing a height field over the triangles is generated. Contrary to earlier approaches we are able to automatically generate these wrinkle textures without user interaction, making it applicable for automatic simulation systems like TüTex. Furthermore, we are the first who combined these textures with a collision-free subdivision step in order to realize displacement mapping. TüTex and Virtual Try-On: All the collision detection and post-processing methods presented in this thesis were included into the cloth simulation engine TüTex. With Virtual Try-On the first system was realized that allows the physically-based simulation of garment based on CAD cloth patterns, physical material parameters and 3D body scans. To be able to easily set up scenes for cloth simulation and to rapidly visualize the achieved results, TüTex was integrated into Alias Maya by converting it to a plugin.

Das Dokument erscheint in: