Integrating multiple viewpoints for articulated scene model aquisition

Abstract

In this paper we present a method to generate an Articulated Scene Model for a system's current view, which allows to integrate multiple egocentric models previously gathered from different viewpoints. The approach is designed to build up separate representations for the static, movable and dynamic parts of an observed scene. In order to make already gathered information available for subsequent viewpoints of the same location, a merging algorithm is needed that considers view-dependent aspects like occlusion and limitations of the view frustum. We show in our experiments that the proposed algorithm correctly merges multiple scene models and can be applied profitably in an integrated vision system for detecting movable objects on a mobile robot. © 2013 Springer-Verlag.