@InProceedings{plagemann08icra,
  title     = {Monocular Range Sensing: A Non-Parametric Learning Approach},
  author    = {Christian Plagemann and Felix Endres and J\"urgen Hess and Cyrill Stachniss and Wolfram Burgard},
  booktitle = {Proc.~of the IEEE Int.~Conf.~on Robotics \& Automation (ICRA)},
  address   = {Pasadena, CA, USA},
  year      = {2008},
  abstract  = {For many applications, mobile robots need to estimate the geometry of their local surrounding area.  To do so, proximity sensor such as laser range finders or sonars are typically employed.  Cameras are a cheap and lightweight alternative to such sensors, but do not offer proximity information directly.  In this paper, we present a novel approach to learning the relationship between range measurements and visual features extracted from a single monocular camera image.  As the learning engine, we apply Gaussian processes, a non-parametric learning technique that not only yields the most likely range prediction corresponding to a certain visual input but also the predictive uncertainty.  This information, in turn, can be utilized in an extended grid-based mapping scheme to update a model of the environment more gently where the predictions are unreliable.  In practical experiments carried out with a mobile robot equipped with an omnidirectional camera system in different environments, we show that our system is able to predict range scans accurate enough to construct maps of the environment.}
}