Evaluation of Recent Approaches to Visual Odometry from RGB-D Images

  • Estimation of camera motion from RGB-D images has been an active research topic in recent years. Several RGB-D visual odometry systems were reported in literature and released under open-source licenses. The objective of this contribution is to evaluate the recently published approaches to motion estimation. A publicly available dataset of RGB-D sequences with precise ground truth data is applied and results are compared and discussed. Experiments on a mobile robot used in the RoboCup@Work league are discussed as well. The system showing the best performance is capable of estimating the motion with drift as small as 1 cms under special conditions, though it has been proven to be robust against shakey motion and moderately non-static scenes.

Export metadata

Additional Services

Share in Twitter Search Google Scholar Availability
Document Type:Conference Object
Author:Sergey Alexandrov, Rainer Herpers
Parent Title (English):Behnke, Veloso et al. (Eds.): RoboCup 2013: Robot World Cup XVII. Lecture Notes in Computer Science, Vol 8371
First Page:444
Last Page:455
Publication year:2014
Departments, institutes and facilities:Fachbereich Informatik
Dewey Decimal Classification (DDC):0 Informatik, Informationswissenschaft, allgemeine Werke / 00 Informatik, Wissen, Systeme / 004 Datenverarbeitung; Informatik
Entry in this database:2015/04/02