The problem of transformation invariant object recognition is considered. W
e develop a projective transformation invariant representation for both sce
ne and model which facilitates an attributed relational graph object matchi
ng based only on unary and binary relations. The unary and binary measureme
nts used for matching are derived from sets of reference points such as cor
ners and bi-tangent points which are stable under the various transformatio
ns considered. Each set of reference points is used to generate a distinct
barycentric coordinate basis system associated with one node of the object
graph representation. We show that barycentric coordinates of the reference
image points can be made invariant under any arbitrary projective transfor
mation. The conditions that must hold for a basis to be valid are stated. W
e illustrate the construction of the barycentric coordinate systems for the
affine and perspective transformations. For the object and scene represent
ation we use the barycentric coordinates of the reference points generating
the barycentric coordinate system, together with auxiliary measurements su
ch as colour and texture as the node's unary measurements. For binary measu
rements we use the product of the barycentric coordinate system for one nod
e with the inverse of the barycentric coordinate system associated with ano
ther node. The unary and binary relations provide an orthogonal decompositi
on of the shape being matched. They are used in a relaxation process to det
ect instances of objects consistent with a given model. We demonstrate the
proposed methodology of projective transformation invariant object represen
tation on several examples. First we illustrate the stability of the shape
representation in terms of unary relations both visually and numerically. W
e then experimentally demonstrated the invariance of binary relations on a
star-like object. We show experimentally that the binary relations derived
are invariant. The final example demonstrates the proposed approach as a to
ol for 3D object recognition. The aim is to recognize 3D objects in terms o
f planar faces. A hexagonal model shape is hypothesized in the image. The o
nly instance of the hypothesized model is successfully recovered. (C) 1999
Elsevier Science B.V. All rights reserved.