The generalized Hough transform (GHT) is a powerful method for recogni
zing arbitrary shapes as long as the correct match accounts for both m
uch of the model and much of the sensory object. For moderate levels o
f occlusion, however, the GHT can hypothesize many false solutions. In
this paper, we present an improved two-stage GHT procedure for the re
cognition of overlapping objects. Each boundary point in the image is
described by three features including the concavity, radius and normal
direction of the curve segment in the neighborhood of the point. The
first stage of the voting process determines the rotational angle of t
he sensory object with respect to the model by matching those points t
hat have the same concavity and radii. The second stage then determine
s the centroid of the sensory object by matching those points that hav
e the same concavity, radii and rotational angles. The three point fea
tures remove the false contribution of votes in the vote generation ph
ase. Experimental results have shown that the proposed algorithm works
well for complex objects under severely overlapping conditions. (C) 1
997 Elsevier Science B.V.