《RS-SLAM RANSAC Sampling for Visual Fast SLAM》.pdfVIP

2011 IEEE/RSJ International Conference on Intelligent Robots and Systems September 25-30, 2011. San Francisco, CA, USA RS-SLAM: RANSAC Sampling for Visual FastSLAM Gim Hee Lee, Friedrich Fraundorfer, and Marc Pollefeys ¨ Computer Vision and Geometry Laboratory, Department of Computer Science, ETH Zurich, Switzerland glee@student.ethz.ch, fraundorfer@inf.ethz.ch, marc.pollefeys@inf.ethz.ch Abstract— In this paper, we present our RS-SLAM algorithm model, he introduced a relocalizer in his PTAM framework for monocular camera where the proposal distribution is [6] which relocalizes the camera during localization failures. derived from the 5-point RANSAC algorithm and image feature measurement uncertainties instead of using the easily violated In this paper, we present our RS-SLAM which is a constant velocity model. We propose to do another RANSAC new monocular SLAM framework that combines the 5- sampling within all the inliers that have the best RANSAC point RANSAC [7], [8] and FastSLAM [9] algorithms. We score to check for inlier misclassifications in the original cor- propose a visual FastSLAM based framework which makes respondences and use all the hypotheses generated from these consensus sets in the proposal distribution. This is to mitigate use of the 5-point RANSAC algorithm and image feature data association errors (inlier misclassifications) caused by the measurement uncertainties as the proposal distribution for the observation that the consensus set from RANSAC that yields the particles during the prediction step instead of using