Home Summary Discussion Support Code Documentation Publications Wishlist Data Sets		A tutorial on how to sequentially match a data set In this tutorial we are going to use the hannover data set for illustration. You can download the hannover data set from
slam6D -s 1 -e 65 -r 10 -i 100 -d 75 --epsICP=0.00001 home/dat/hannover1/

-s 1

This parameter is for specifying where the scan should start so the data set registration will start from the first scan.

-e 65

This parameter is for specifying where the scan should end so the registration is going to end after the 65th scan.

-r 10

This is the point reduction parameter. It will reduce the points to a voxel size of 10.

-i 100

This parameter specifies that there should be 100 iterations in the sequential matching.

-d 75

This parameter specifies that the maximal distance of points that may form corresponding point pairs is 75.

--epsICP=0.00001

Each iteration of the ICP algorithms minimizes the error between all point pairs. This parameter specifies one of the stopping criteria of the algorithm. If in subsequent iterations the difference of average distances between the point pairs after matching falls below this value, the matching process is terminated. The sequential matching of the data set contains some error. This parameter specifies the stopping criteria of the iteration. In this case, if the difference between two points is smaller than 0.00001, the iteration will stop.

/home/dat/hannover1/

This is the destination path for the folder containing the scans to be registered. This may vary depending on where you save the hannover1 data set.

To visualise or animate the data set you can refer to the show tutorials on this page. An example command for using show to graphically display the sequentially matched data set is:

show -s 1 -e 65

This will display the sequentially matched data set starting from the 1st scan and ending on the 65th scan.

Below are shots of the output generated by show for the sequentially matched data set.





It can be seen from the second shot that there is an overlapping of the wall. This error can be removed by using simultaneous matching
to match the data set.

Related Publications

• Andreas Nüchter, Jan Elseberg, Peter Schneider, and Dietrich Paulus. Study of Parameterizations for the Rigid Body Transformations of The Scan Registration Problem, Journal Computer Vision and Image Understanding (CVIU), Elsevier Science, Volume 114, Issue 8, pp. 963-980, ISSN 1077-3142, August 2010. [Get Paper (PDF)] [Elsevier Link with supplementary content].

• Deyuan Qiu, Stefan May, and Andreas Nüchter. GPU-accelerated Nearest Neighbor Search for 3D Registration. In Proceedings of the 7th International Conference on Computer Vision Systems (ICVS '09). LNCS 5815, Spinger ISBN 978-3-642-04666-7, pages 194-203, Lìege Belgium, October 2009. [Get Paper (PDF)]

• Oliver Wulf, Andreas Nüchter, Joachim Hertzberg, and Bernardo Wagner. Benchmarking Urban Six-Degree-of-Freedom Simultaneous Localization and Mapping. Journal of Field Robotics (JFR), Wiley & Son, ISSN 1556-4959, Volume 25, Issue 3, pages 148 - 163, March, 2008, [Get Paper] [Get Videos].

• Andreas Nüchter, Kai Lingemann, and Joachim Hertzberg. Cached k-d tree search for ICP algorithms. In Proceedings of the 6th IEEE International Conference on Recent Advances in 3D Digital Imaging and Modeling (3DIM '07), IEEE Computer Society Press, ISBN 0-7695-2939-9, pages 419 - 426, Montreal, Canada, August 2007, [Get Paper (PDF)] [HTML Version].

• Andreas Nüchter, Kai Lingemann, Joachim Hertzberg, and Hartmut Surmann, 6D SLAM - 3D Mapping Outdoor Environments Journal of Field Robotics (JFR), Special Issue on Quantitative Performance Evaluation of Robotic and Intelligent Systems, Wiley & Son, ISSN 1556-4959, Volume 24, Issue 8-9, pages 699 - 722, August - September, 2007, [Get Paper].