| Title | Globally consistent registration of time-of-flight camera images |
| Publication Type | Master Thesis |
| Authors | Artner, P. |
| Supervised By | Swadzba, A., and S. Wachsmuth |
| Academic Department | Faculty of Technology |
| Publisher | Bielefeld University |
| Year of Publication | 2009 |
| Abstract | To be able to successfully navigate, any mobile robot needs a representation
of it’s environment. Nowadays, mobile robots use simultaneously localisation and mapping (SLAM) techniques and 3D scanners to build 3D maps of the surrounding area.
This task is done by registering many overlapping 3D scans into one consistent representation of the surrounding environment. The quality of the resulting representation is crucial for its main use: navigation and interaction.
The state-of-the-art algorithm for 3D cloud registration is the iterative
closest/corresponding point (ICP) algorithm. Since 1992 the basic algorithm
has undergone and is still undergoing many enhancements. One such refinement, named approximated ICP (aprxICP), was presented in 2009 by Nüchter and Elseberg. Nüchter and Elseberg evaluated the aprxICP on laser scan data acquired by a mobile robot exploring an urban outdoor area.
This thesis attempts to adjust the aprxICP for indoor time-of-flight (ToF)
camera data. To compensate the more inexact ToF camera data an extension
to meta scan matching is tested. |
| Keywords | ICP, registration, time-of-flight sensor |
| Citation Key | 2290 |
