Abstract

In this paper, an experimental model-based vision system recognizing silhouettes of 3-D objects is described. The system uses local constraints to prune a tree of pairings between image features and model features, then performs a complete verification for each remaining hypothesis. Recognition based on local constraints has been reported by Grimson and Lozano-Perez for matching 2-D measurements to 2-D models, and for matching 3-D measurements to 3-D models [1]. The system described in this paper matches 2-D silhouette measurements to 3-D object models.

© 1987 Optical Society of America