International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
   
  
Figure 3: Union of objects according to a theme classification: 
DS2 (up) and DS1 (down) 
  
Figure 4: Match between the new objects of DSI (the thick 
polygon) and DS2 (red-filled polygon) 
Apparently, matching the newly created objects is much easier 
and intuitive (Figure 4). 
The second algorithm was tested with data sets DS1 and DS3, 
i.c. they were used to create data set DS2. Recall the beginning 
of this Section about the description of the data sets. DSI and 
DS2 have the same scale but different theme and DS2 and DS3 
have the same theme, but different scale. Thus the aim is to 
create objects that have the higher resolution of DSI (scale 
1:1000) and the theme classification of DS2. As mentioned 
before, the three data sets are maintained separately. From 
company point of view, this process will greatly reduce efforts 
and time to create DS2 (suppose DS1 and DS3 already exist and 
DS2 does not yet exist). 
Figure 5 shows a new object as a result of the executed script. 
Shown are the set of objects of DS1 that intersect with one 
object of DS3 and have similar themes. After aggregating the 
entire set of selected objects, a new object is created that 
follows the outer boundary of the aggregation. The difference 
between the original object of DS2 (the thick polygon polygon) 
and the newly created object is minimal. 
Figure 5: Object of DS2 (the thick polygon) generated from 
DS1 (the objects inside) and DS3. 
The tests have shown that the procedure is very successful. 
Several factors have to be taken into consideration while 
generating new objects: ontology schema, precision and 
accuracy of maps and gross errors. We expect that in some 
cases manual intervention will still be necessary. 
The complexity of the process as mentioned above varies with 
respect to the data sets used. For example several more 
iterations are needed for the link between DSI and DS2 
compared to the link between DS2 and DS3. The number of 
objects in the DSI is rather large and there are still quite many 
other errors (e.g. self-overlapping polygons, gaps, wrong layers, 
etc). As it was shown, many objects of DS1 (parts of rivers, or 
other objects) can be merged first (using appropriate conditions) 
and after that linked to DS2 objects. The link can be further 
controlled with respect to the classification of the two types of 
objects. 
6 CONCLUSIONS 
Although the optimal case would be to have one DBMS 
representation of a real world object, multiple representations 
(based on different themes or scale) exist and will continue to 
exist. Support for multiple representations in DBMSs is 
indispensable with respect to the growing role of DBMSs in the 
new generation GIS architecture. 
In this context, the functionality of DBMS is critical. Two 
aspects of multiple representations in DBMS were explored in 
this paper: 1. automatically establish links between different 
data sets and 2. automatically generation of low-resolution data 
sets from high-resolution data sets. Two algorithms utilising 
Oracle Spatial functions were developed and tested with three 
different data sets. 
The functionality currently offered by Oracle Spatial allows 
establishing a link between the three data sets from the case 
study. From this study it can be concluded that the results 
obtained from overlapping geometries are better than comparing 
themes. The complexity of the process also varies with respect 
to the data sets used. 
The aggregation of several objects into a new object according 
to thematic characteristics is a relatively simple and 
straightforward process, which can be easily completed utilising 
available spatial functions in the DBMS. This can be used as a 
practical solution for comparing objects from different 
applications. The geometries of the objects can be aggregated 
according to a hierarchical classification up to a level at which 
the aggregated objects have the same meaning. Combinations of 
theme aggregations with geometry overlap gives the freedom to 
develop procedures with a diverse complexity. The creation of 
new data sets is an example of such an elaborated procedure. 
Data consistency and integrity would be significantly improved 
if multiple representations were organised in one virtual DBMS. 
Different representations of the same real-world object can be 
stored in a distributed DBMS environment together with the 
226 
Inter 
  
integ 
repre 
be a 
diffe 
autoi 
Furtl 
follo 
ACK 
This 
Tran: 
Susta 
Tech 
REF 
Bink 
Orac 
6-9, : 
Franl 
carto 
grapt 
OGC 
Speci 
Sche 
Docu 
OGC 
speci 
Proje 
Mass 
Friis- 
Manz 
Butte 
Repre 
Repo 
Buffa 
Devo 
Multi