A new technique to recognise 3D free-form objects via registration is
proposed. This technique attempts to register a free-form surface, rep
resented by a set of 2 1/2D sensed data points, to the model surface,
represented by another set of 2 1/2D model data points, without prior
knowledge of correspondence or vie;tv points between the two point set
s. With an initial assumption that the sensed surface be part of a mor
e complete model surface, the algorithm begins by selecting three disp
ersed, reliable points on the sensed surface. To find the three corres
ponding model points, the method uses the principal curvatures and the
Darboux frames to restrict the search over the model space. Invariabl
y, many possible model 3-tuples will be found. For each hypothesized m
odel 3-tuple, the transformation to match the sensed 3-tuple to the mo
del 3 tuple can be determined. A heuristic search is proposed to singl
e out the optimal transformation in low order time. For realistic obje
ct recognition or registration, where the two range images are often e
xtracted from different view points of the model, the earlier assumpti
on that the sensed surface be part of a more complete model surface ca
nnot be relied on. With this, the sensed 3-tuple must be chosen such t
hat the three sensed points lie on the common region visible to both t
he sensed and model views. We propose an algorithm to select a minimal
non-redundant set of 3-tuples such that at least one of the S-tuples
will lie on the overlap. Applying the previous algorithm to each 3-tup
le within this set, the optimal transformation can be determined. Expe
riments using data obtained from a range finder have indicated fast re
gistration for relatively complex test cases. If the optimal registrat
ions between the sensed data (candidate) and each of a set of model da
ta are found, then, for 3D object recognition purposes, the minimal be
st fit error can be used as the decision rule.