Three dimensional object recognition using axial feature signatures

A method for recognizing 3-D polyhedra using a passive vision system is presented. Object surfaces are modeled as an aggregation of triangular surfaces, each triangular surface in turn being represented by its vertices. Features are extracted from the two-dimensional perspective projection silhouette as the object is rotated about a predetermined axis. These features, as determined by a complete revolution around the object about this axis, constitute the axial feature signatures (AFS) for the modeled polyhedra solid. Features include the area, perimeter, and other 2-D characteristics of the silhouette of the object. The process of recognition is to match the AFS of modeled objects in the database to the signature of a polyhedron under visual scrutiny.<>