[Zurück]


Vorträge und Posterpräsentationen (mit Tagungsband-Eintrag):

S. Chroust, M. Vincze:
"Detection of Motion Discontinuities in Visual Servoing";
Vortrag: 26th Workshop of the Austrian Association for Pattern Recognition (ÖAGM/AAPR), Graz; 10.09.2002 - 11.09.2002; in: "Vision with Non-Traditional Sensors", Österreichische Computer Gesellschaft, 160 (2002), ISBN: 3-85403-160-2; S. 133 - 140.



Kurzfassung englisch:
The main drawback of visual sensors is the delay time, which is often handled by predic-
tive algorithms, namely after Kalman. These algorithms show a bad prediction behaviour for
discontinuities in the target motion. In this paper a method is introduced to detect such dis-
continuities by supervising the prediction error with a Prediction Monitor (PM) to react faster
and with less oscillations. This new prediction algorithm is able to achieve a good prediction
quality for smooth and discontinuous motions. The new approach is evaluated using a pan/tilt
unit to track a coloured object and the results are compared to an Adaptive Kalman Filter.
The improvements in the experiments are 25 - 40 % depending on the criteria and the motion
pattern.


Elektronische Version der Publikation:
http://www.infa.tuwien.ac.at/groups/robtec/Veroeff/ChroustVincze.pdf


Erstellt aus der Publikationsdatenbank der Technischen Universitšt Wien.