hand gestures and pointing. Some systems required
the user to wear a special glove or magnetic
sensors [I, 2, 3]. Others using image processing
have required calibration for each user's
individual hand shape and posture [4].
We have developed a stereo vision pointing
system as an input device for a robot manipulator,
to provide a novel and convenient means for
tho operator to specify points for pick-and-place
operations. We use active contour techniques [5]
to track a hand in a pointing gesture, with conventional
monochrome cameras and fairly modest
image-processing hardware.
A single view of a pointing hand is ambiguous:
its distance from the camera cannot be determined,
and the slant of its orientation is hard
to measure. Stereo views constrain the indicated
point to a line in space, passing through the fingertip
in the direction of pointing. We employ
a ground-plane const,raint to determine the single
point on a tnl,lc-lop corresponding 1.0 each
gesture. The use of such a constraint effectively
reduces the problem to a two-dimensional one.