The localization of mobile systems with video data is a challenging field in robotic vision research. Apart from artificial environmental support technologies like GPS localization, a selfsufficient visual system is desirable. In this work, we introduce a new heuristic approach to outdoor localization in a scenario where no odometry readings are available. In an earlier work, we employed SIFT features and a common particle filter method in the scenario. A modification of Particle Swarm Optimization, a popular optimization technique especially in dynamically changing environments, is developed and fit to the localization problem, including self-adaptive mechanisms. The new method obtains similar or better localization results in our experiments, while requiring a fraction of SIFT comparisons of the standard method, indicating an all-over speed-up by 25%.
[pdf]
@inproceedings{Kron07Loc,
author = {Kronfeld, Marcel and Weiss, Christian and Zell, Andreas},
title = {Swarm-supported Outdoor Localization with Sparse Visual Data},
booktitle = {3rd European Conference on Mobile Robots (ECMR 2007)},
year = {2007},
pages = {259--264},
address = {Freiburg, Germany},
abstract = {The localization of mobile systems with video data is a challenging
field in robotic vision research. Apart from artificial environmental
support technologies like GPS localization, a selfsufficient visual
system is desirable. In this work, we introduce a new heuristic approach
to outdoor localization in a scenario where no odometry readings
are available. In an earlier work, we employed SIFT features and
a common particle filter method in the scenario. A modification of
Particle Swarm Optimization, a popular optimization technique especially
in dynamically changing environments, is developed and fit to the
localization problem, including self-adaptive mechanisms. The new
method obtains similar or better localization results in our experiments,
while requiring a fraction of SIFT comparisons of the standard method,
indicating an all-over speed-up by 25%.},
url = {http://www.cogsys.cs.uni-tuebingen.de/publikationen/2007/Kron07SwarmSuppFinal.pdf}
}