Part identification using a superquadric representation /

Bibliographic Details
Main Author: Messimer, Sherri Lynn
Other Authors: Blank, L. T. (degree committee member.), Friesen, D. K. (degree committee member.), Malaye, C. O. (degree committee member.), Mitta, D. A. (degree committee member.)
Format: Thesis Book
Language:English
Published: 1989.
Subjects:
Online Access:Link to OAKTrust copy
Description
Abstract:This dissertation presents a methodology for the recognition of two-dimensional complex parts. A complex part is defined as a collection of rigid subparts which do not undergo independent transformations. The representation scheme utilized is discussed as well as the strategies used in this matching process. The representation scheme used to describe a complex part is a relational graph constructed of superquadric primitives. In this graph, a subpart of the object is described by a node in the graph, and the relationship between nodes is given by labeled edges. Three tasks are performed in order to ascertain the identity of the unidentified complex parts. First, the values of the parameters defining the superquadric shape must be determined from the input image. A nonlinear regression technique, based on Marquardt's method, is used to estimate the shape, size, and orientation parameters for each subpart belonging to the complex part. These values are stored on the nodes of the graph structure and are used to facilitate the matching procedure. Second, local matching procedures are used to match nodes (or subparts) contained in the digitized image to those stored in the database. This is achieved by checking whether the digitized subpart image lies inside an error zone generated by each database subpart. The error zone is calculated using the values of the superquadric shape and size parameters found using the nonlinear estimation technique. A level of confidence is assigned to each local match based on this error zone calculation. All subparts are treated in this same manner until the list of subparts is exhausted..
Item Description:Typescript (photocopy).
Vita.
"Major subject: Industrial engineering."
Physical Description:xii, 132 leaves : illustrations ; 29 cm
Bibliography:Includes bibliographical references.