![](bg8.png)
0.05. For the plane detection, the minimum count of
points needed to constitute a plane was set to a value
between 200 and 2000, depending on the density of
the dataset. The radius of the box filter for bitmap
smoothing was chosen such that it corresponds to ap-
proximately 20 cm in point cloud coordinates. Visi-
bility tests between pairs of points were computed on
the GPU using OpenCL. The experiments were con-
ducted on a 3.5 GHz Intel Core i7 CPU and a GeForce
GTX 670 GPU with 4 GB of memory. To obtain syn-
thetic data, we implemented a virtual laser scanner
which simulates the scanning process within 3D CAD
building models.
10 CONCLUSIONS AND FUTURE
WORK
We presented a method for the extraction of structural
building descriptions using 3D point cloud scans as
the input data. Our method was evaluated using syn-
thetic and real-world data, showing the feasibility of
our approach. In most cases, the algorithm produced
satisfactory results, yielding useful semantic repre-
sentations of buildings for applications like naviga-
tion in point clouds or structural queries.
The usage of different visibility functionals as
well as an EM-formulation to determine the param-
eters of the used normal distributions separately for
each room label could help to improve the robustness
of our method. Performance in the presence of non-
convex rooms could possibly be improved by using
a measure for (potentially indirect) reachability be-
tween points instead of visibility tests. Also, limiting
the tests to a more local scope may help to overcome
the identified problems. The output of our approach
could be further enriched by more attributes, for in-
stance by applying methods for the analysis of the in-
dividual room point sets after segmentation.
ACKNOWLEDGEMENTS
We would like to thank Henrik Leander Evers for the
scans of Kronborg Castle, Denmark, and the Faculty
of Architecture and Landscape Sciences of Leibniz
University Hannover for providing the 3D building
models that were used for generating the synthetic
data. This work was partially funded by the Euro-
pean Community’s Seventh Framework Programme
(FP7/2007-2013) under grant agreement no. 600908
(DURAARK) 2013-2016.
REFERENCES
Ahmed, S., Liwicki, M., Weber, M., and Dengel, A.
(2012). Automatic room detection and room label-
ing from architectural floor plans. In 10th IAPR In-
ternational Workshop on Document Analysis Systems
(DAS-2012), pages 339–343. IEEE.
Bokeloh, M., Berner, A., Wand, M., Seidel, H.-P., and
Schilling, A. (2009). Symmetry detection using line
features. Computer Graphics Forum (EG 2009),
28(2):697–706.
Dempster, A. P., Laird, N. M., and Rubin, D. B. (1977).
Maximum likelihood from incomplete data via the em
algorithm. JOURNAL OF THE ROYAL STATISTICAL
SOCIETY, SERIES B, 39(1):1–38.
Duda, R., Hart, P., and Stork, D. (2001). Pattern Classifica-
tion. Wiley.
Kim, Y. M., Mitra, N. J., Yan, D.-M., and Guibas, L. (2012).
Acquiring 3d indoor environments with variability
and repetition. ACM TOG, 31(6):138:1–138:11.
Koppula, H. S., Anand, A., Joachims, T., and Saxena, A.
(2011). Semantic labeling of 3d point clouds for in-
door scenes. In NIPS 2011, pages 244–252.
Langenhan, C., Weber, M., Liwicki, M., Dengel, A., and
F., P. (2012). Topological query on semantic building
models using ontology and graph theory. In EG-ICE.
TUM.
Langenhan, C., Weber, M., Liwicki, M., Petzold, F., and
Dengel, A. (2013). Graph-based retrieval of build-
ing information models for supporting the early de-
sign stages. Advanced Engineering Informatics.
Nan, L., Xie, K., and Sharf, A. (2012). A search-classify ap-
proach for cluttered indoor scene understanding. ACM
TOG, 31(6):137:1–137:10.
Rusu, R. B., Marton, Z. C., Blodow, N., Dolha, M., and
Beetz, M. (2008). Towards 3D point cloud based ob-
ject maps for household environments. Robotics and
Autonomous Systems, 56(11):927941.
Schnabel, R., Wahl, R., and Klein, R. (2007). Efficient
ransac for point-cloud shape detection. Computer
Graphics Forum, 26(2):214–226.
Schnabel, R., Wessel, R., Wahl, R., and Klein, R. (2008).
Shape recognition in 3d point-clouds. In WSCG 2008.
UNION Agency-Science Press.
Verma, V., Kumar, R., and Hsu, S. (2006). 3d building de-
tection and modeling from aerial lidar data. CVPR
2012, 2:2213–2220.
Wessel, R., Bl
¨
umel, I., and Klein, R. (2008). The room
connectivity graph: Shape retrieval in the architectural
domain. In WSCG 2008. UNION Agency-Science
Press.
Xiao, J. and Furukawa, Y. (2012). Reconstructing the
world’s museums. In Proceedings of the 12th Euro-
pean Conference on Computer Vision, ECCV ’12.
AutomaticGenerationofStructuralBuildingDescriptionsfrom3DPointCloudScans
127