We adress the problem of localizing a mobile agent in a known environment, where we estimate the position and orientation based on an omnidirectional vision sensor. In general, this environment can be given as a metric map or as in our case, as a database of sensor measurements with associated metric pose information, in contrast to purely topological approaches, where the associated information is of semantic nature.
Using our novel regression-based mapping and localization scheme, the training data, consisting of position and appearance information, can be acquired with minimal effort. By relying on odometry readings and by incorporating the propagated uncertainty into the model, a probabilistic map is created that allows for efficient localization from very low-resolution images.
Using the regression approach, it is possible to avoid discrete decisions in the algorithm. We do not establish explicit correspondences in neither the spatial nor the image domain. Instead, a measure of distances is learned in image space which, together with estimated uncertainties about the position information, leads to improved localization estimates, as an alternative to explicit loop closure.
Localization results using perfect reference poses
Localization results using odometry-based mapping
[Schairer-2011-Visual] Timo Schairer and Benjamin Huhle and Andreas Schilling Wolfgang Straßer, Visual Mapping with Uncertainty for Correspondence-free Localization using Gaussian Process Regression, in: Proc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), 2011
[Huhle-2010-Learning] Benjamin Huhle and Timo Schairer and Andreas Schilling and Wolfgang Straßer, Learning to Localize using Gaussian Process Regression on Omnidirectional Image Data, in: IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), 2010
[Schairer-2010-Application] Timo Schairer and Benjamin Huhle and Wolfgang Straßer, Application of Particle Filters to Vision-Based Orientation Estimation using Harmonic Analysis, in: IEEE International Conference on Robotics and Automation (ICRA), 2010
[Schairer-2010-SIFTVs.] Timo Schairer and Sebastian Herholz and Benjamin Huhle and Wolfgang Straßer, SIFT vs. SOFT - A Comparison Of Feature And Correlation Based Rotation Estimation For Panoramic Images, in: 3DTV CON - The True Vision, 2010
[Huhle-2009-Normalized] Benjamin Huhle and Timo Schairer and Wolfgang Straßer, Normalized Cross-Correlation using SOFT, in: Proc. International Workshop on Local and Non-Local Approximation in Image Processing (LNLA), 2009
[Schairer-2009-Increased] Timo Schairer and Benjamin Huhle and Wolfgang Straßer, Increased Accuracy Orientation Estimation from Omnidirectional Images using the Spherical Fourier Transform, in: 3DTV CON - The True Vision, 2009
(For links to the papers, please goto "Department/Publications" or visit the personal pages of the respective authors)