ISPRS. Vol.34, Part 2W2. ^Dynamic and Multi-Dimensional GISL Bangkok May 23-25 2001_ 
ISPRS, Vol.34, 
48 
Server-side Client-side 
Fig.3: Client-Server 3D Internet GIS 
3.2 Storing large data set and Internet Query analysis 
The multiple and large amount of building data storing in server 
database require the data management to be powerful and 
efficient. The traditional RDBMS has several weaknesses in 
managing 3D geometry and attribute data, especially in hybrid 
data structure. Now, some commercial databases such as 
Oracle 8i can handle Object-Relation data and provide binary 
large object (BLOB) type, permitting image and spatial data 
management. They can also construct lager Internet database 
and provide many data interface to access the database 
Considering the entire data of all buildings such as millions of 
thousands buildings in a city, it is impossible to keep all the 
data in memory of a workstation for visualization and analysis, 
so a well-organized data structure and fast data retrieval 
mechanism have to be provided 
In order to efficiently manage 3D building data in database, the 
BLOB can be extended to store building model. For the 
purpose of query and analysis, the operation of writing and 
reading BLOB from database must be defined, at the same 
time, the index of BLOB also must be provided to fasten the 
data retrieval. The R-tree structure is proved to be a useful 
index. Compared with other spatial data structure such as 
Quad-trees or Oct-trees, it is obvious that R-tree is much better 
suited for the organization of overlapping object, the rectangle 
bounding boxes of 2D or 3D CAD model of building, 
buildingblocks and large urban units (Gruber, 1998). According 
to the R-tree, the block of divide building can be a district or 
several buildings For the purpose of visualization, different 
abstract layer of LC model data and R-tree index require to be 
stored in database. According to LC model, the 3D building 
object can be depicted as follow: 
BuildingObject=( Layers, BodyObjectArray, RoofObjectArray. 
R-tree index). 
BodyObjectArray=( Bodyl ,Body2, ...Bodyn). 
RoofObjectArray=( Roof 1,Roof2,.....Roofn). 
Body=( PointArray, Height), Roof=(BoundaryArray, 
FeaturePointArray). 
Using the R-tree index to organize the building data on 
server-side, the JDBC is used to implement the query the 
building attribute and SQL analysis. Using JDBC-ODBC bridge, 
the user can communicate between the client-side and the 
server-side to query the building attribute and some query 
analysis (See Figure 4). In order to improve the efficiency of 
query on Internet, the middleware on Web server, which is 
used to connect the Geodatabase, is needed. Figure 5 shows 
the result of the query of building attribute 
3. EXPERIi 
A system prot 
developed, a 
model througt 
construct the 
index at datat 
client-side do\ 
while the city 
The user can