In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
326 
3.2 Test Results 
Figure 4 and 5 show the results of performance comparisons of 
the four types. For non-spatial queries, Type 4’s retrieval time is 
fastest in performance followed by 3, 2 and 1 in order. As the 
number of 3D objects increases, query performance for all four 
models decreases. While Type 4 shows the smallest increasing 
rate, Type 1 shows the greatest. As can be seen, with less than 
1.000 units, four models show differences up to 6 seconds, 
implying that all four models are suitable for visualization for 
the size of urban model including hundreds of building objects. 
However, with over 1,000 objects, the retrieval performances 
show significant gaps among the models, implying that Type 4 
is the only alternative for visualization. 
In the test for the non-spatial queries for 100,000 units in the 
PostGIS, the retrieval time of each model shows as follows; 
Type 1 - 369.3 seconds, Type 2 - 202.4 seconds, Type 3 - 
143.0 seconds, Type 4-19.1 seconds. In the spatial query tests, 
Type 3’s retrieval time is fastest in performance followed by 1, 
4 and 2. All four models show steady increase as the number of 
objects increases. The reason that spatial queries show less time 
than non-spatial queries is we experimented using a portion (20 
x 20) from entire objects, while non-spatial treats all the objects. 
The number of objects from the range query was around 10. 
From the result, we can see that spatial range queries can be 
applied to all four models for retrieving relatively small number 
of objects from less than 100,000 units, since the gaps between 
them area 5 seconds at most. 
* 
3.3 Visualization 
VRML supports diverse base solid features and allows the 
combination of multimedia such as animation or sound with 
applications, which makes visualizing 3D models in VRML 
relatively easy. Low-level graphic libraries such as OpenGL and 
Direct X can describe objects more in detail. In the test, we used 
real 3D building models of around 500 stored in the PostGIS. 
Although both techniques showed similar quality, VRML 
showed significant decrease in refreshing speed as the number 
of objects in the display increases. Thus, it is viewed that 
OpenGL is more suitable for 3D visualization applications 
although development with OpenGL requires more time than 
VRML (Figure 6 and 7). 
Figure 6. An example of visualization test using VRML 
Figure 7. An example of visualization test using OpenGL 
4. CONCLUDING REMARKS 
Although 3D models are getting used increasingly in many 
areas including architecture, urban planning and environmental 
analysis, they mostly have been used as visualization purposes 
without using topological structure or semantic information. 3D 
topology models have been studied over the last decade and 
some of them were connected with DBMS-based 
implementations. However, we haven’t found any attempts for 
comparing the models in the viewpoint of performances as of 
now. In this study, we categorized previously proposed 3D 
models and compared them including our own model using 
different queries to the SDBMS. We tested the retrieval time by 
non-spatial and spatial queries and suggested the most 
computationally favorable models for LADAR simulations. 
Also, we compared visualization performances between VRML 
and OpenGL. The results imply that choosing a 3D modeling 
should be done according to the problem requirement types. 
Type 1 is shown to be proper for visualization, while type 3 is 
for spatial queries. However, types of 3D model may 
additionally be determined by the level of details of the classes 
where the required semantic information resides. For example, 
when modeling indoor spaces, we many need a more 
decomposed model like Type 1 which fully implements the 
topological relationships between objects. 
ACKNOWLEDGEMENTS 
This research was supported by the Seoul R&BD Program 
(10561), Korea. 
REFERENCES 
Ellul, C. and Haklay, M., 2008. Using a B-rep structure to query 
9-intersection topological relationships in 3D GIS - reviewing 
the approach and improving performance. In J. Lee and S. 
Zlatanova, (eds.), 3D Geo-information Sciences, Springer- 
Verlag, Berlin, pp. 127 -151. 
Groger, G., Reuter, M. and Plümer, L., 2004. Representation of 
a 3-D city model in spatial object-relational databases. In: The 
20th Congress of International Society for Photogrammetry 
and Remote Sensing, Istanbul, Turkey. 
Kolbe, T., 2008. Representing and exchanging 3D city models 
with CityGML. In J. Lee and S. Zlatanova, (eds.), 3D Geo 
information Sciences, Springer-Verlag, Berlin, pp. 15 -31. 
Stadler, A. and Kolbe, T., 2007. Spatio-semantic coherence in 
the integration of 3D city models. In: The 5th International 
ISPRS Symposium on Spatial Data Quality ISSDQ 2007, in 
Enschede. 
Stoter, J. and van Oosterom, P., 2002. Incorporating 3D geo 
objects into a 2D geo-DBMS. In: ACSM-ASPRS 2002 Annual 
Conference. Washington DC. 
Stoter, J. and Zlatanova, S., 2003. Visualising and editing of 3D 
objects organised in a DBMS. In: EUROSDR Workshop : 
Rendering and Visualisation, Enschede, The Netherlands, pp. 
14-29. 
Zlatanova, S., 2000, 3D GIS for urban development, Ph.D. 
thesis, Institute for Computer Graphics and Vision, Graz 
University of Technology, Austria, ITC, the Netherlands.