Object Representation Using Appearance-Based Parts and Relations

摘要

The recognition of general three-dimensional objects in cluttered scenes remains a challenging problem. In particular, the design of a good representation that is suitable for modeling large numbers of generic objects, and is also robust to occlusion, has been a stumbling block to achieving success. In this paper, we propose a representation using appearance-based parts and relations to overcome these problems. Appearance-based parts and relations are defined in terms of closed regions and unions of these regions, respectively. The regions are segmented using the MDL principle; their appearance is obtained from collections of images and compactly represented by parametric manifolds in the two eigenspaces spanned by the parts and the relations. Qualitative and quantitative experiments illustrating the potential of the representation for successful object recognition in the presence of clutter and occlusion are presented.