Automated surface acquisition using range cameras.

Format:

Book

Thesis/Dissertation

Author/Creator:

Pito, Richard Anthony.

Contributor:

Bajcsy, Ruzena, advisor.

University of Pennsylvania.

Language:

English

Subjects (All):

Computer science.

Remote sensing.

Electrical engineering.

0544.

0799.

0984.

Penn dissertations--Computer and information science.

Computer and information science--Penn dissertations.

Local Subjects:

Penn dissertations--Computer and information science.

Computer and information science--Penn dissertations.

0544.

0799.

0984.

Physical Description:

185 pages

Contained In:

Dissertation Abstracts International 58-03B.

System Details:

Mode of access: World Wide Web.

text file

Summary:

This work addresses the problem of using a range camera to automatically acquire accurate surface descriptions of complex rigid objects whose geometry and topology are a priori unknown. Since a range scanner cannot sample through an object, multiple range images must be acquired from different vantage points, registered and integrated to form a complete model. The next best view (NBV) problem is to automatically determine a position and orientation for the range scanner, subject to several constraints, from which it will scan into some unseen portion(s) of the viewing volume. The registration problem is to accurately recreate the relative positions and orientations of the scanned surfaces of the object in the range data. The integration problem is to robustly merge noisy and potentially overlapping and intersecting range data into a consistent manifold surface representation. By considering each problem in context of an automated as opposed to a manual surface acquisition system, not only is each further qualified but their interdependence is firmly established. By making an accurate accounting of the physics of range data acquisition, raw range data is enriched with information which is not only used to eliminate noise in the range data itself but which is crucial in solving the NBV and integration problems. A general solution to the NBV problem is presented which utilizes a novel data structure, positional space, capable of simultaneously representing those areas of the viewing volume which must and which can be scanned. The algorithm is capable of considering many thousands of potential camera positions by performing costly visibility operations a fixed number of times, independent of the number of potential camera positions. The causes of registration failure are analyzed to produce a registration aid which can be used to accurately and robustly register even non-overlapping range data. Finally, a robust surface based integration algorithm is presented which is based on a neighborhood metric defined using the sampling physics of range data acquisition. The algorithm is guaranteed to produce a manifold and plausible surface. These solutions are integrated into an automated surface acquisition system which is demonstrated on complex real objects.

Notes:

Thesis (Ph.D. in Computer and Information Science) -- University of Pennsylvania, 1997.

Source: Dissertation Abstracts International, Volume: 58-03, Section: B, page: 1375.

Supervisor: Ruzena Bajcsy.

Local Notes:

School code: 0175.

ISBN:

9780591362718

Access Restriction:

Restricted for use by site license.