Authors:
Frauke Wübbold
and
Bernardo Wagner
Affiliation:
Leibniz Universität Hannover, Germany
Keyword(s):
Object Classification, Segmentation, Feedback, 3D Point Cloud, 3D Shape.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Mobile Robots and Autonomous Systems
;
Perception and Awareness
;
Robotics and Automation
Abstract:
Limited knowledge and limited deduction abilities are among the main restraints of autonomous robots for acting truly autonomously. This especially becomes obvious in the area of object recognition and classification, where many methods rely on knowledge teached manually in a prior setup step. Self-generating this knowledge from environment perception with a set of rules would significantly increase the robots autonomy as well as supersede manual training effort. In this paper, we propose a novel approach to rule-based classification for 3D point clouds by means of object shape, which additionally overcomes typical problems from a separate prior segmentation by integrating classification feedback into the segmentation process. Although it is still in its conceptual state, we explain in detail why we consider this approach to be very promising.