Large-scale localization using only a camera

The automation of a robot to measure ice sheets is necessary for CReSIS. The approach described in this paper is to use a single camera for localization for use in many types of large-scale environments. In order to operate in large-scale environments, a hybrid map approach has been used; both topological and geometric maps. An appearance-based approach is used for recognizing the different locations. In order to simplify testing, all experiments were performed offline and the topological and geometric testing was performed separately. The results of the testing showed that 94% of the images were localized correctly in the testing environment. The system typically required at least two images to solve the global localization problem, and around three images to solve the kidnapped robot problem. These images are localized to within 1 foot and 45 degrees of the actual position and orientation.