MULTIPLE REPRESENTATIONS IN DBMS: TWO ALGORITHMS 
Jantien Stoter^ and Sisi Zlatanova® 
? International Institute for Geo-Information Science and Earth Observation (ITC), Hengelosestraat 99, Enschede, The Netherlands, 
stoter(@ite.nl 
^ Delft University of Technology, Jaffalaan 9, Delft, The Netherlands, s.zlatanova(otb.tudelft.nl 
Commission 1V/3 
KEY WORDS: Databases, Generalization, Framework, Data structures, Representation, Retrieval, Modelling, Automation 
ABSTRACT: 
Spatial data sets are created to provide information for specific applications. These sets are representations of real world objects and 
are each collected by specific organisations for specific purposes. The objects of interests are modelled in a way appropriate for the 
application and therefore the data sets are a subjective selection of real world objects. Consequently, different representations of the 
same objects can be found in a variety of data sets. Dealing with several representations is related to two major research domains: 
efficient structuring and intelligent generalisation. 
This paper addresses these two issues with respect to the functionality offered by spatial DBMS. Growing number of mainstream 
DBMS have been offering management of spatial objects. The number of implemented spatial operations increases as well. Many of 
these operations (or combinations of them) can be already successfully used to manage multi-resolution data. The paper discusses 
possibilities for data structuring (using the spatial data types offered by DBMS), algorithms for automatic linking of different 
representations and generation of new representations out existing ones. The algorithms are tested in a case study. 
% 
1 INTRODUCTION 
A lot of research has been already conducted on multiple 
representations and generalisation related to spatial DBMS: 
Buttenfield and DeLotto, 1989; Devogele et al., 1996; Friis- 
Christensen et al., 2002; Grefen and Widom, 1997; Jones et al., 
1996; Li and McLeod, 1992; Sheth and Larson, 1990; 
Spaccapietra et al., 1999), which is a clear indication for the 
significance of the issue and the need for support of multiple 
representations in DBMSs. 
Two general principles for data organisation can be 
distinguished: single-resolution management (one real world 
object is translated into one instance in the database) and 
multiple resolution (one real world object has several 
representations in the database). Proposed frameworks by Frank 
and Timpf, 1994 and Vangenot, 1998 are examples of the first 
principle. This approach is very appropriate for modelling new 
data sets. 
However, maintenance of several representations in one 
database is unavoidable in many cases. Multiple representations 
of real world objects already exist in many organisations. A 
typical example are the different representations of the same 
real world objects for different applications. Single 
representations are dependent on the subjective view of the user 
who has modelled the representation. This view is related to 
different aspects such as scale (also dependent on the amount of 
details to be represented), generalisation criteria, theme and 
time (Zhou and Jones, 2003). To be able to exchange such 
representations and to use the representations from one 
application in another application, a correspondence between 
the different data sets has to be established. Due to lack of 
efficient methods to establish a link between such 
representations (thematically and geometrically), very often a 
new process is initiated for data collection, modelling and 
updating of the same objects. 
Here, we investigate the functionality presently offered by 
spatial DBMS to resolve two specific issues related to different 
geometries of one object existing in different data sets, Le.: 
e Possibilities for establishing a link between different 
representations of objects 
e Automatic generation of low-resolution 
representations from high-resolution representations 
In this paper both issues are discussed in the context of an 
object-relational DBMS. The paper is organised in four 
sections. Section 2 describes the functionalities offered by 
mainstream DBMSs to support multiple representations in 
DBMSs. Section 3 describes approaches for linking different 
geometries, while section 4 focuses on automatic generation of 
low-resolution representations. Section 5 reports the results of a 
case study utilising developed functions and scripts. The case 
study is carried out within Oracle Spatial 91. 
The paper concludes on the usability of spatial functions for 
multi-resolution management at DBMS level and outlines 
further research topics 
2 DATA STRUCTURING OF MULTIPLE 
REPRESENTATIONS IN DBMS 
DBMS plays an important role in the new generation GIS 
architecture. The algorithms to interrelate different data sets and 
to create low-resolution data from high-resolution data arc 
based on core DBMS functionality. Therefore, the 
functionalities available in DBMSs that can support modelling 
of multiple representations are presented first (see also Stoter 
and Zlatanova, 2003). 
Interna. 
21 5G 
Mainstt 
have in 
called ' 
Feature 
implem 
Data T: 
updatin 
ADTs 
technol 
in geot 
geomet 
section 
To spe: 
data set 
Most I 
index. 
An im] 
validati 
relation 
stored : 
area, € 
intersec 
of cooi 
multipli 
geomet 
small p 
reality. 
were c 
process 
2:2 sS 
Spatial 
DBMS: 
betweei 
sets. 
The OC 
describ 
support 
geomet 
betwee 
nine-int 
each sp 
The bc 
interior 
end poi 
its peri 
object I 
points 
relation 
with the 
‘covere 
within 1 
otherwi 
first ob 
bounda 
returns 
In Ingr 
Oracle, 
and toy 
functioi 
Oracle 
analysis 
Oracle | 
topolog 
speciali