Feature selection for appearance-based robot localization
Ben Kröse, Nikos Vlassis, R. Bunschoten, Yoichi Motomura · 2000
In this paper we present a method for an appearance based modeling of the environment of the robot. We use a Markov localization procedure in which the model gives a probabilistic relation between linear image features and the position of the robot. We developed a method to select those linear features which are best for localization. We show how feature selection influence the localization performance using data acquired within the Real World Computing Partnership project.