ial data. 
al world 
| several 
field of 
ce level. 
1l as the 
set into 
  
  
  
  
s by 
ne with 
search on 
but at the 
roblem is 
automatic 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
matching tool for our research that gives us the possibility to 
match the data comfortably by hand and use these matchings as 
an input for the integration of data schemas. 
This work is part of the Nexus project (Nexus 04). In the Nexus 
project, we are developing an open platform for all possible 
types of mobile, location-based information systems. In order to 
realize a generic approach in Nexus, different data providers 
have to be able to integrate their data into the Nexus world 
model. For this reason, a schema integration takes place that 
maps the object classes of existing data schemas from data 
providers onto the classes of the Nexus schema. At the moment 
this process is done manually. In this paper we show how this 
can be done in an automatic way. The paper first gives an 
overview on related work. In section 3, it is discussed how 
spatial databases can be related. Section 4 comprises a detailed 
explanation of the realization of our approach. 
2. RELATED WORK 
The topic of spatial data integration is very much related to the 
research arcas listed below. Some of their aspects will be briefly 
presented in the following section 
un 
The notion of the research presented in this paper has already 
been addressed by (Uitermark 1996): “Geographic Data set 
integration (or map integration) is the process of establishing 
relationships between corresponding object instances in 
different, autonomously produced, geographic data sets of a 
certain region. The purpose of geographic data set integration is 
to share information between different geographic information 
sources”. 
2.1 Matching and conflation 
Concerning the matching of spatial objects, the basic idea is to 
express and to evaluate the similarity of spatial features. If a 
certain degree of similarity can be detected, two features can be 
assigned to each other. (Bruns & Egenhofer 1996) have adopted 
this basic assumption and count the steps that have to be taken 
to transform one representation into another representation. The 
number of steps can then be interpreted as a similarity measure. 
A fundamental, line-based matching approach for street network 
data has been presented by (Walter and Fritsch 1999). In a first 
step, the algorithm finds all potential correspondencies of 
topologically connected line elements in two source data sets by 
performing a buffer operation. The matching candidates are 
stored in a list. This list is ambiguous and typically contains a 
large amount of n:m matching pairs. Then, unlikely matching 
pairs are identified and eliminated using relational parameters 
like topologic information and feature-based parameters like 
line angles. The result is a smaller but still ambiguous list with 
potential matching pairs. These matching pairs are evaluated 
with a merit function in order to compute a unique combination 
of matching pairs which represents the solution of the matching 
problem. This is a combinatorial problem which is solved with 
an A* algorithm. 
A point-based matching method was proposed, for example, in 
(Bofinger 2001). The algorithm developed here is based on the 
idea of describing intersections of streets, i.e. nodes of a street 
network, by an explicitly defined code. The code consists of 
point coordinates, abbreviations and names of incident streets 
and the number of linked edges. For cach intersection, such a 
153 
code is created. By comparing the codes of the intersections 
within different data sets and by assigning the intersections with 
the most similar codes to each other, references can be derived. 
The problem of conflation is for example being tackled by 
(Cobb et al. 1998). The merging process is defined here as 
"feature deconfliction", where all parts of a matched feature pair 
are unified into a single "better" feature. The conflation 
algorithm has to decide, which properties are preserved in the 
resulting instance. In their approach, the authors are also taking 
into account the data quality information of the corresponding 
instances. 
2.2 Semantic data integration and ontologies 
According to (Uitermark et al. 1999), semantic integration can 
be understood as a communication process since two partners 
who want to communicate have to have the same understanding 
of the objects they are talking about. 
In the database domain, some work has been done regarding 
schema matching by (Do and Rahm 2002), where schemas are 
compared using parameters like element names, data types or 
further structural information. In the field of GIS, a lot of 
different approaches have been carried out using ontologies. 
Ontologies can be defined as formalized specifications of 
concepts about objects of the real world from a certain 
application perspective (Gruber 93). Whereas database schemas 
require a digital representation, ontologies are just abstract 
views on the semantics of things. There is only one ontology for 
an object in a certain application domain, but there can be 
multiple database representations (Fonseca et al. 2002). 
Consequently, concerning schema integration, two cases have to 
be considered (Hakimpour and Timpf 2001): 
I. Database schemas arc based on the same ontology: 
only synonyms and homonyms have to be detected to 
perform an integration. 
2. Database schemas are based on different ontologies 
(from different application domains): a common 
ontology has to be created by detecting the 
similarities of the source ontologies. 
The authors are presenting a formalism for the representation of 
ontologies, the so-called Description Logic (DL). Each user 
community can define its perception of an object using DL and 
then different ontologies can be merged. 
Another example on how to integrate different semantics of 
spatial data is provided by (Bishr et al. 1999). The approach 
consists of two components, the Semantic Wrapper and the 
Semantic Mapper. Objects of different spatial databases are 
wrapped by the Semantic Wrapper and have to conform to a 
predefined interface so that they can be recognised by the 
Semantic Mapper. This interface is specific for a certain 
application domain like transportation, topography, etc. On the 
level of the Semantic Mapper, the semantics of two objects can 
be compared and the schematic and semantic differences 
between them can be resolved. 
2.3 Standardization 
The question of interoperability of GIS is mainly addressed by 
the OpenGIS Consortium (OGC 2004) and the Technical 
Commission 211 of the International Standards Organization 
(ISO-TC211 2004). Both institutions are closely linked.