A Fast Visual Odometry and Mapping System for RGB-D Cameras
Authors:
- Bruno M.F. SilvaBSBruno M.F. Silva
- Luiz M.G. GoncalvesLuiz M.G. Goncalves
- 1 October 2014
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 55-60
Abstract
The introduction of low cost range sensing devices such as RGB-D cameras allows applications for Robotics to exploit novel and real-time capabilities. One such application is Visual Odometry, a module responsible to use the synchronized color/depth streams captured by this class of sensors to estimate the position and orientation of a robot at the same time that a map representation of the environment is built. Aiming to localize robots in a fast and efficient way, we design a Visual Odometry system for RGB-D sensors that allows real-time (approximately 25 Hz) camera pose estimation despite the fact that no specialized hardware (such as modern GPUs) is employed. Experiments carried out on publicly available benchmark and datasets demonstrate the usefulness of the method, which achieved localization accuracy superior to the state-of-the-art RGB-D SLAM algorithm.Keywords
This publication has 27 references indexed in Scilit:
- Robust odometry estimation for RGB-D cameras2013
- Robust real-time visual odometry for dense RGB-D mapping2013
- RGB-D mapping: Using Kinect-style depth cameras for dense 3D modeling of indoor environmentsThe International Journal of Robotics Research, 2012
- G2o: A general framework for graph optimization2011
- Three‐dimensional mapping with time‐of‐flight camerasJournal of Field Robotics, 2009
- Distinctive Image Features from Scale-Invariant KeypointsInternational Journal of Computer Vision, 2004
- Bundle Adjustment — A Modern Synthesis2000
- A method for registration of 3-D shapesIEEE Transactions on Pattern Analysis and Machine Intelligence, 1992
- Least-squares estimation of transformation parameters between two point patternsIEEE Transactions on Pattern Analysis and Machine Intelligence, 1991
- Random sample consensusCommunications of the ACM, 1981