Tree of Words for Visual Loop Closure Detection in Urban SLAM
2008 (English)In: Proceedings of the '08 Australasian Conference on Robotics and Automation, 2008, 102- p.Conference paper (Refereed)
This paper introduces vision based loop closure detection in Simultaneous Localisation And Mapping (SLAM) using Tree of Words. The loop closure performance in a complex urban environment is examined and an additional feature is suggested for safer matching. A SLAM ground experiment in an urban area is performed using Tree of Words, a delayed state information ﬁlter and planar laser scans for relative pose estimation. Results show that a good map estimation using our vision based loop closure detection can be obtained in near real, yet constant, time. It is shown that an odometry supported recall rate of almost 70% can be obtained with a false detection rate of about 0.01%.
Place, publisher, year, edition, pages
2008. 102- p.
Tree of words, Simultaneous localisation and mapping
National CategoryEngineering and Technology Control Engineering
IdentifiersURN: urn:nbn:se:liu:diva-44676Local ID: 77269ISBN: 978-0-646-50643-2OAI: oai:DiVA.org:liu-44676DiVA: diva2:265538
'08 Australasian Conference on Robotics and Automation, Canberra, Australia, December, 2008