Isolated 3D object recognition through next view planning

Abstract

Most object recognition systems use information from a single image of an object. In many cases, a single view may not contain su cient features to recognize the object unambiguously. Hence, more than one view is necessary. With an active sensor, the recognition process therefore involves identi cation of a view of an object and if necessary, planning the next view. This paper presents a new on-line recognition scheme based on next view planning for the identi cation of an isolated 3D object using simple features. The scheme uses a probabilistic reasoning framework for recognition and planning. We present a knowledge representation scheme which encodes both feature-based information about objects in the model base as well as the uncertainty in the recognition process. This scheme is used both in the probability calculations as well as in planning the next view. The recognition scheme is on-line wherein past observations guide the planning process. Results clearly demonstrate its e ectiveness for a reasonably complex experimental set. 1