Universität Bielefeld - Sonderforschungsbereich 360
Hand-eye calibration:
neuronal and analytical approaches
Dr.
Guo-Qing Wei (DLR, Oberpfaffenhofen)
Dienstag, 9. Dezember 1997
14 - 16 Uhr
Hörsaal 3
Abstract:
In this talk, I consider the hand-eye coordination problem
in the multisensory case (cameras and laser range finders).
First, a method is proposed to learn the direct transformation
from sensory data to hand motions. By training a multi-layer
perceptron for one homing position, the network can be modified
such that new homing positions can be achieved without re-relearning.
As an alternative to the neural network method, an analytical
method is also proposed to identify the physical parameters of the
sensors necessary for hand-eye coordination. Closed form solutions
and conditions for a unique solution are derived in terms
controlled motion sequences. Experimental results are presented,
and advantages and disadvantages of each method are compared.
Anke Weinberger, 1997-11-10, 1997-11-18