[Zurück]

M. Vincze, S. Chroust, P. Gemeiner, U. Mühlmann, A. Pinz:
"Egomotion Estimation in an Unknown Environment from Vision and Inertial Sensor Data for Mobile Robots and Augmented Reality Applications";
Vortrag: International Workshop on Robotics in Alpe-Adria-Danube Region, Brno; 02.06.2004 - 05.06.2004; in: "Proceedings of RAAD'04", (2004), S. 154 - 159.

For mobile robotics or head gears in augmented reality (AR) applications it is essential to continuously
estimate the egomotion and the structure of the environment. In robotics this is needed for navigation,
in AR to overlay virtual information correctly. The applications demand full 3D pose estimation of the
egomotion at a, preferably, high sampling rate and with sufficient precision for navigation and virtual
display. The ultimate goal is to be able to investigate new environments. This paper will present the
system developed in the SmartTracking project, which achieves these demands by integrating vision and
inertial sensors. While vision sensors are known to loose targets during fast motion, inertial sensors can
estimate rapid motions but suffer from drift during slow motions. Hence, these two complimentary sensor
types are combined in one estimation scheme. Estimation is based on corner features, which can be found
in many environments. From a single known starting position, the system can move into an unknown
environment. It uses specially developed CMOS cameras that operate at a rate of 2000 windows per
second. The vision and inertial data are fused with an extended Kalman filter. The filter is designed to
handle asynchronous input from these two sensors, which typically operate at different and possibly varying
rates. Additionally to the 3D trajectory of the egomotion, the structure of the environment is built up
from the corner features detected. The 3D pose estimation scheme is given in full detail, demonstrated in a
set-up with an active sensor head and compared to ground truth obtained by a reference tracking system.

http://aleph.ub.tuwien.ac.at/F?base=tuw01&func=find-c&ccl_term=AC04967676