Cs. Chua et R. Jarvis, POINT SIGNATURES - A NEW REPRESENTATION FOR 3D OBJECT RECOGNITION, International journal of computer vision, 25(1), 1997, pp. 63-85
Citations number
27
Categorie Soggetti
Computer Sciences, Special Topics","Computer Science Artificial Intelligence
Few systems capable of recognizing complex objects with free-form (scu
lptured) surfaces have been developed. The apparent lack of success is
mainly due to the lack of a competent modelling scheme for representi
ng such complex objects. In this paper, a new form of point representa
tion for describing 3D free-form surfaces is proposed. This representa
tion, which we call the point signature, serves to describe the struct
ural neighbourhood of a point in a more complete manner than just usin
g the 3D coordinates of the point. Being invariant to rotation and tra
nslation, the point signature can be used directly to hypothesize the
correspondence to model points with similar signatures. Recognition is
achieved by matching the signatures of data points representing the s
ensed surface to the signatures of data points representing the model
surface. The use of point signatures is not restricted to the recognit
ion of a single-object scene to a small library of models. Instead, it
can be extended naturally to the recognition of scenes containing mul
tiple partially-overlapping objects (which may also be juxtaposed with
each other) against a large model library. No preliminary phase of se
gmenting the scene into the component objects is required. In searchin
g for the appropriate candidate model, recognition need not proceed in
a linear order which can become prohibitive for a large model library
. For a given scene, signatures are extracted at arbitrarily spaced se
ed points. Each of these signatures is used to vote for models that co
ntain points having similar signatures. Inappropriate models with low
votes can be rejected while the remaining candidate models are ordered
according to the votes they received. In this way, efficient verifica
tion of the hypothesized candidates can proceed by testing the most li
kely model first. Experiments using real data obtained from a range fi
nder have shown fast recognition from a library of fifteen models whos
e complexities vary from that of simple piecewise quadric shapes to co
mplicated face masks. Results from the recognition of both single-obje
ct and multiple-object scenes are presented.