THESIS
2017
xiii, 126 pages : illustrations ; 30 cm
Abstract
With the improvement of CAD/CG (computer-aided design and computer graphics)
technology and the development of VR (virtual reality) and AR (augmented reality) techniques,
more and more real-world items surrounding people have been digitalized and displayed in the
virtual world as well as people themselves. Among all the items, garment product is an essential
one to be recreated for its close relationship with human body, but it is labor- and time-intensive
to construct and simulate the models due to the need of pattern making background, large
garment style/shape variation and flexible object nature. Meanwhile, current e-commerce
platforms have a requirement to convert real garment products into 3D (three-dimensional)
models followed by providing customers with not only fine ren...[
Read more ]
With the improvement of CAD/CG (computer-aided design and computer graphics)
technology and the development of VR (virtual reality) and AR (augmented reality) techniques,
more and more real-world items surrounding people have been digitalized and displayed in the
virtual world as well as people themselves. Among all the items, garment product is an essential
one to be recreated for its close relationship with human body, but it is labor- and time-intensive
to construct and simulate the models due to the need of pattern making background, large
garment style/shape variation and flexible object nature. Meanwhile, current e-commerce
platforms have a requirement to convert real garment products into 3D (three-dimensional)
models followed by providing customers with not only fine rendering effect but also online
virtual dressing service. Although state-of-art CG techniques can satisfactorily achieve lifelike
garment modeling and rendering in movies, they are still not capable to handle enormous online
virtual dressing and individual garment fitting demand. Based on these reasons, a systematic
framework and related techniques are proposed in this dissertation to improve the efficiency of
garment transposition, garment customization based on feature-based model matching method.
More specifically, a scenario from real data collection, human and garment modeling, to
database construction and virtual dressing is firstly introduced, based on which a
comprehensive literature review on classic and recent cloth simulation, garment modeling and
model matching is given to analyze and compare existed methods on solving garment
transposition and customization problems. To break a bottleneck in virtual dressing industry, a
solution workflow is introduced to address garment transposition and the manipulation
problems in 3D garment customization. The foundation of proposed solution is feature-based
model matching method by which different garment models and human models are linked
together and parameterized using same set of features. Each process of transposition and
manipulation acts as an operation executed from object-level down to feature-level followed by
geometry-level. Detailed algorithms such as human model cross-parameterization and feature-based
automatic garment segmentation are designed and the rules to control model matching
are also given. According to the relationship obtained from garment and human model matching,
several garment model manipulation and deformation tools acting on parameter domain are
developed to adjust model shape and meet fitting requirement. Moreover, some detail
techniques such as geometric collision analysis, strain analysis and limiting catering for
proposed garment transposition method are also enhanced based on existed works to refine
deformed models being more visually plausible.
Benefitting from the features of proposed solution and detailed techniques, an example
database consisting of dressed human characters can be built for virtual garment oriented
application such as online virtual dressing, individual garment customization or even real-time
animation.
Post a Comment