3D object identification with color and curvature signatures

In this paper we describe a model-based object identification system. Given a set of 3D objects and a scene containing one or more of these objects, the system identifies which objects appear in the scene by matching surface signatures. Surface signatures are feature vectors that reflect the probability of occurrence of the features for a given surface. Two types of surface signatures are employed; curvature signatures and spectral (i.e. color) signatures. Furthermore, the system employs an inexpensive acquisition setup consisting of a single CCD camera and two light sources. The system has been tested on 95 observed surfaces and 77 objects with varying degrees of curvature and color with good results. ( 1999 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.