Degree Type

Dissertation

Date of Award

2011

Degree Name

Doctor of Philosophy

Department

Mechanical Engineering

First Advisor

James Oliver

Second Advisor

Eliot Winer

Abstract

Engineering product design is an information intensive decision-making

process that consists of several phases including design specification

definition, design concepts generation, detailed design and analysis,

and manufacturing. Usually, generating geometry models for

visualization is a big challenge for early stage conceptual design.

Complexity of existing computer aided design packages constrains

participation of people with various backgrounds in the design

process. In addition, many design processes do not take advantage of

the rich amount of legacy information available for new concepts

creation.

The research presented here explores the use of advanced graphical

techniques to quickly and efficiently merge legacy information with

new design concepts to rapidly create new conceptual product designs.

3D mesh metamorphosis framework "3DMeshMorpher" was created to

construct new models by navigating in a shape-space of registered

design models. The framework is composed of: i) a fast spherical

parameterization method to map a geometric model (genus-0) onto a unit

sphere; ii) a geometric feature identification and picking technique

based on 3D skeleton extraction; and iii) a LOD controllable 3D

remeshing scheme with spherical mesh subdivision based on the

developedspherical parameterization. This efficient software framework

enables designers to create numerous geometric concepts in real time

with a simple graphical user interface.

The spherical parameterization method is focused on closed genus-zero

meshes. It is based upon barycentric coordinates with convex boundary.

Unlike most existing similar approaches which deal with each vertex in

the mesh equally, the method developed in this research focuses

primarily on resolving overlapping areas, which helps speed the

parameterization process. The algorithm starts by normalizing the

source mesh onto a unit sphere and followed by some initial relaxation

via Gauss-Seidel iterations. Due to its emphasis on solving only

challenging overlapping regions, this parameterization process is much

faster than existing spherical mapping methods.

To ensure the correspondence of features from different models, we

introduce a skeleton based feature identification and picking method

for features alignment. Unlike traditional methods that align single

point for each feature, this method can provide alignments for

complete feature areas. This could help users to create more

reasonable intermediate morphing results with preserved topological

features. This skeleton featuring framework could potentially be

extended to automatic features alignment for geometries with similar

topologies. The skeleton extracted could also be applied for other

applications such as skeleton-based animations.

The 3D remeshing algorithm with spherical mesh subdivision is

developed to generate a common connectivity for different mesh models.

This method is derived from the concept of spherical mesh subdivision.

The local recursive subdivision can be set to match the desired LOD

(level of details) for source spherical mesh. Such LOD is controllable

and this allows various outputs with different resolutions. Such

recursive subdivision then follows by a triangular correction process

which ensures valid triangulations for the remeshing. And the final

mesh merging and reconstruction process produces the remeshing model

with desired LOD specified from user. Usually the final merged model

contains all the geometric details from each model with reasonable

amount of vertices, unlike other existing methods that result in big

amount of vertices in the merged model. Such multi-resolution outputs

with controllable LOD could also be applied in various other computer

graphics applications such as computer games.

Copyright Owner

Ruqin Zhang

Language

en

Date Available

2012-04-30

File Format

application/pdf

File Size

123 pages

Share

COinS