167 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B6b. Beijing 2008 
the groundwork for designing and building many different 3D 
LBS applications or upgrading the ones from existing 2D LBS. 
3.1 Basic concepts 
Many existing 2D LBS are using the WFS to provide diverse 
datasets because of its powerful data processing and analyzing 
capabilities. Those LBS providers also expect to keep the same 
framework when upgrading the current 2D LBS to 3D LBS in 
order to reduce the costs and make the maintenance task simple. 
As described before, CityGML as a special case of the GML for 
representing 3D data can be supported by the WFS, and thus the 
combination of the CityGML and WFS is a good option to 
provide 3D LBS. Therefore, the way of how to make use of 
existing 3D datasets and publish them as the CityGML through 
the WFS needs to be investigated. 
There are large amounts of existing BIM data representing 
detailed 3D buildings with high LoDs containing such as floors, 
ceilings, walls and doors objects, which belong to the LoD4 in 
the CityGML. The BIM data is normally stored as CAD data 
format like IFC and DXF. To employ those existing 3D datasets 
into LBS or GIS, the interoperability issue has to be aware. An 
idea is to load those datasets into a spatial database for further 
data handling. However, there is currently no standard encoding 
for storing 3D geometry in the database, which hinders the 
WFS in serving 3D data (Müller and Curtis, 2005). Müller and 
Curtis proposed to model the complex 3D geometry types into 
separate tables which aggregate other geometries via table joins. 
Thus, the 3D geometry can be stored either as build-in database 
types or further aggregations through table joins. 
After putting the 3D BIM datasets into the database, the WFS 
can be created to connect with the database and deliver the LBS 
client applications the GML data with 3D encoding (CityGML) 
via a standard interface. For the LBS client, it needs not know 
the internal changes of the LBS system. 
3.2 Methodology 
Based on the basic concepts, the workflow for setting up a 3D 
LBS system is illustrated in Figure 3. In general, the source data 
of BIM format is firstly converted into the GML data by the 
data translation and transformation tool named Feature 
Manipulation Engine (FME) 2007 from Safe Software. Then 
the GML data is loaded into Oracle 10g database through 
Snowflake GoLoader. After that, a standardized WFS server 
based on the Oracle database is built through Snowflake 
GoPublisher toolset. Finally, through the WFS interface the 
client can obtain the CityGML data used for different 3D LBS 
applications. 
Figure 3. A prototype of 3D LBS using CityGML and WFS 
To testify the proposed workflow, a simple building is firstly 
created using AutoDesk Architectural Desktop in the DWG 
format. The DWG file is then converted to a GML3 document 
by FME. However, FME has not yet fully supported 3D proxy 
objects of the DWG format, so the achieved GML document 
needs to be restructured in order to keep the semantic 
information, e.g. which polygon geometries belong to a 
“window” object. After that, Snowflake GoLoader and 
GoPublisher tools are employed to fulfil the following tasks. 
Finally, we can retrieve a CityGML file containing the same 
spatio-semantic content as the original DWG file through 
standard WFS requests. It then can be visualized in the 3D 
GML Aristoteles viewer as shown in Figure 4. 
The results have also shown that the functions of an existing 
WFS server are not affected with the newly added 3D datasets, 
because the interface to the LBS client is not changed. Thus, a 
2D LBS system may be upgraded to a 3D LBS system in a 
similar way. 
Figure 4. The small building as the CityGML data visualized in 
3D GML Aristoteles viewer 
4. CONCLUSIONS AND OUTLOOK 
In this paper, we have discussed the important issues when 
upgrading 2D LBS to 3D LBS based on several core 
components: mobile devices, location sensing techniques, data 
standards and standardized geospatial web services. However, 
the introduction of 3D LBS encounters a number of challenging 
issues. Because of the high demands of 3D data processing and 
visualization, 3D LBS need the improvements on all the above 
aspects. For example, mobile devices with better performance 
are required; more accuracy and coverage positioning service is 
preferred, and thus the hybrid model of different location 
sensing techniques should be considered; suitable spatio- 
semantic 3D data formats like CityGML can be the role to serve 
3D LBS; existing standardized geospatial web services like 
WFS can be upgraded to meet complex 3D data requirements 
without changing its standard interface. 
A prototype has been given to show the possibility of upgrading 
an existing 2D LBS system to the 3D LBS system. The 
CityGML and WFS are the key roles to fulfil the tasks. Though 
several software tools can greatly help to implement the