Drexel University Home Pagewww.drexel.edu DREXEL UNIVERSITY LIBRARIES HOMEPAGE >>
iDEA DREXEL ARCHIVES >>

iDEA: Drexel E-repository and Archives > Drexel Theses and Dissertations > Drexel Theses and Dissertations > Scale-dependent/invariant local 3D geometric features and shape descriptors

Please use this identifier to cite or link to this item: http://hdl.handle.net/1860/2791

Title: Scale-dependent/invariant local 3D geometric features and shape descriptors
Authors: Novatnack, John
Keywords: Computer Science;Computer graphics;Three-dimensional display systems
Issue Date: 25-Jun-2008
Abstract: The quality and abundance of three-dimensional geometric data is rapidly increasing with the drastic improvement in the cost and effectiveness of 3D acquisition hardware. In fact, threedimensional geometric data already plays a central role in many computer vision and computer graphics applications such as autonomous vehicle navigation, 3D object recognition and the computerbased preservation of cultural artifacts. Despite the increasing relevance and importance of geometric data, current techniques of processing the data have neglected to explicitly model and exploit a significant source of information of the data - the scale variability of the local geometric structures. In this thesis we overcome the limitation of past techniques with a comprehensive framework of modeling the scale-variations in local geometric structures, effectively adding an additional dimension to geometric data. To accomplish this we derive the geometric scale-space, a representation of local geometric structures at various degrees of scale. This representation enables us to define scale-dependent geometric feature detectors, such as corners and edges, that determine not only the location of salient geometric features, but also their relative scales. The augmentation of a geometric feature with its intrinsic scale enables us to define scale-dependent/invariant local shape descriptors that together form both a hierarchical and scale-invariant representation of the local geometric structures of a 3D shape. We derive and present the theory of these methods and also demonstrate their effectiveness for the purposes of robust 3D feature detection and fully automatic range image registration.
URI: http://hdl.handle.net/1860/2791
Appears in Collections:Drexel Theses and Dissertations

Files in This Item:

File Description SizeFormat
Novatnack_John.pdf2.13 MBAdobe PDFView/Open
View Statistics

Items in iDEA are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Valid XHTML 1.0! iDEA Software Copyright © 2002-2010  Duraspace - Feedback