@inproceedings {pub3498,
	title = {Condition Invariant Visual Localization Using Slow Feature Analysis},
	author = {Muhammad  Haris AND Benjamin Metka AND Mathias Franzius AND Ute Bauer-Wersing},
	year = {2017},
	month = {September},
	abstract = {In outdoor scenarios varying environmental conditions like
seasonal, weather and lighting effects have a strong impact on the ap-
pearance which often prevents successful localization. A spatial represen-
tation of the environment can be learned by applying unsupervised Slow
Feature Analysis (SFA) directly to images captured by a mobile robot.
However, effects that change on a slower or equal timescale than the
robot{\textquoteright}s position during learning will be encoded in the resulting repre-
sentations and thus affect spatial coding. In this work we use recordings
from a simulator along the same trajectory, each in a different condition,
which allows to change the perceived image statistics for improved con-
dition invariance. Experiments demonstrate an improvement of spatial
coding even for few training sets.},
	publisher = {Frank-Michael Schleif},
	booktitle = {New Challenges in Neural Computation (NC2)},
	editor = {Thomas Villmann and Frank-Michael Schleif},
	city = {Basel},
	series = {Machine Learning Reports}
}