POSE VERIFICATION AS AN OPTIMAL ASSIGNMENT PROBLEM

In this paper we propose a feature comparison based technique for pose verification for 3-D object recognition and localization from range images. The proposed technique treats the pose verification problem as an optimal assignment problem and could be looked upon as an alternative to the more conventional pose verification technique based on straightforward depth comparison. As is brought out by the experimental results, the proposed technique shows greater efficiency and robustness as compared to the conventional technique. Although the proposed technique is based on dihedral junctions as features for recognition and localization, it could be easily extended to include other feature types.