(Inzwischen sind beide Adressen veraltet. Aktuelle Adresse:
In this paper a new method of modeling 3D-objects for the interpretation of range images is described.
In range data it is possible to detect surface patches, edges and 3D-vertices by local and therefore fast operations. Recognition of an object is done by associating these structures with one element of a set of possible objects. For that it is necessary to use an adequate description of objects which supports fast matching. In this paper a model is presented, which represents an object by surfaces, edges, vertices and neighborhood relationships. The representation of the objects is restricted on those surfaces, which are easy to detect by the used range sensor.
A further requirement is a fast and easy interactive construction of object models which assists human demands and thus reduces the costs of providing and maintenance of the object database.
Keywords: range data, range image, depth data, surface patch, edge, vertex, neighborhood.
This paper is published in the Proceedings of the Second German-Slovenian Workshop on 3-D Scene Acquisition, Modeling and Understanding, held on June 9. 1994 in Ljubljana, Slovenia.
It is also publihed in CIT, Journal of Computing and Information Technology, Vol. 2, No. 4.
This work describes my master thesis, made at the Institute for Robotics and Process Control at the Technical University of Braunschweig. The master thesis was part of a project called Objekterkennung mit einem 3D-Robotersensor (Object recognition with a 3d robot sensor). The thesis is in the library:
I want to thank my advisor Dr. Thomas Stahs for guidance. Also many thanks to Prof. Wahl, the head of the Institute for Robotics and Process Control. And last but not least many thanks to my companions Stephan Lochmann and Christoph Kohnert.