Date of Completion

8-23-2017

Embargo Period

8-23-2017

Keywords

mesh model, CAD, geometry, spectral signature, shape analysis, point cloud

Major Advisor

Horea Ilies

Associate Advisor

Kazem Kazerounian

Associate Advisor

George Lykotrafitis

Associate Advisor

Donald Sheehy

Associate Advisor

Julian Norato

Field of Study

Mechanical Engineering

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

Shape analysis of point cloud surface models produces quality results and avoids the pitfalls of working with mesh models.

Shape analysis is concerned with understanding the shape of models geometrically, topologically, and relationally. Traditionally, shape analysis methods have operated on solid and surface models of objects, especially surface mesh models.

Recent advances in 3D camera technology has driven demand for automatic shape analysis tools.

Devices like the Microsoft Kinect are democratizing 3D sensing and such expansion of what was once an academic and industrial space is making it clear that there is a need for generally-applicable techniques which don't require expert understanding to use.

Mesh model methods require human expertise to ensure suitability for processing. Point cloud models, on the other hand, are the natural output of depth cameras and need no human post-processing to render them amenable to analysis.

This dissertation demonstrates that it is possible to understand shape from point cloud models in ways that don't discard the broad mesh model-based shape analysis literature. Instead, I develop an understanding of how to apply a large class of existing mesh model methods directly on point cloud models without global surface reconstruction. The results obtained by these point cloud model methods are of a quality on par with those obtained by equivalent mesh methods and can additionally avoid entire classes of mesh-specific problems. I also provide a general improvement to the large "spectral" class of shape signature algorithms on any model type.

COinS