We have previously demonstrated that the performance of tracking algor
ithms can be improved by integrating information from multiple cues in
a model-driven Bayesian reasoning framework. Here we extend our work
to active vision tracking, with variable camera geometry. Many existen
t active tracking algorithms avoid the problem of variable camera geom
etry by tracking view independent features, such as corners and lines.
However, the performance of algorithms based on those single features
will greatly deteriorate in the presence of specularities and dense c
lutter. We show, by integrating multiple cues and updating the camera
geometry on-line, that it is possible to track;a complicated object mo
ving arbitrarily in three-dimensional (3D) space. We use a four degree
-of-freedom (4-DoF) binocular camera rig to track three focus features
of an industrial object, whose complete model is known. The camera ge
ometry is updated by using the rig control commands and kinematic mode
l of the stereo head. The extrinsic parameters are further refined by
interpolation from a previously sampled calibration of the head work s
pace. The 2D target position estimates are obtained by a combination o
f blob detection, edge searching and gray-level matching, with the aid
of model geometrical structure projection using current estimates of
camera geometry. The information is represented in the form of a proba
bility density distribution, and propagated in a Bayes Net. The Bayesi
an reasoning that is performed in the 2D image is coupled by the rigid
model geometry constraint in 3D space. An alpha beta filter is used t
o smooth the tracking pursuit and to predict the position of the objec
t in the next iteration of data acquisition. The solution of the inver
se kinematic problem at the predicted position is used to control the
position of the stereo head. Finally, experiments show that the target
undertaking arbitrarily 3D motion can be successfully tracked in the
presence of specularities and dense clutter. (C) 1998 Academic Press.