for the GIS data modelling part. The entity-relationsship 
approach may be seen as a semantic network formalism. 
Analogies to the object-oriented method are given as well. 
An example for a data model of a GIS application based on 
the semantical network approach is given in the following 
chapter (Figure 6). Frames on the other hand are collect- 
ing the various properties of objects in slots. The individual 
values in a slot may be default values, methods, linkages 
to other frames or user given values. Frames are similar to 
complex objects currently investigated in database research 
as part of the structural object-orientation. 
3 APPLICATION OF KNOWLEDGE 
IN GIS 
Different GIS-applications where knowledge integration may be 
very effective should be illustrated now. G. Zhang and J. Tulip 
(1990) are presenting a fuzzy polygon overlay approach solving 
the sliver polygon problem quite reasonable. Figure 5 gives a 
comparison of an exact versus a fuzzy polygon overlay of three 
layers of information. The production rules may look like the 
following simplified statements : 
e If many points resulting from intersection are identical 
within a certain tolerance they are merged together to one 
point. 
e If many edges between points are identical within a certain 
bandwidth, these edges are merged and form one new edge. 
Rule-based approaches may be suitable for raster-vector conver- 
sion and object recognition because of their capability to define 
rules with weak or fuzzy conditions. Usually scanned maps or 
images are no longer showing exact geometric and topological 
features, which causes severe problems for procedural and hard- 
coded approaches for conversion and object recognition. Further 
applications of rule-based methods may be given in rule-based 
assembly of features during data capture, for the production of 
thematic displays and for generalization purposes in GIS. 
A last example illustrates the modelling of the real world struc- 
ture and behaviour based on semantic networks. An extract of 
the real world — a model of fresh water features — is presented in 
figure 6. Semantical networks, frames and scripts are very well 
suited to GIS-data modelling. 
4 REFERENCES 
Bill, R., 1991. Konzepte der Wissens- und Datenreprásentation 
in Geo-Informationssystemen. Proceedings of the 43 rd 
Photogrammetric Week at Stuttgart University. Schriften- 
reihe Heft 15. Institut für Photogrammetrie der Universität 
Stuttgart. pp. 99-113. 
Bill, R., Fritsch, D., 1991. Grundlagen der Geo-Informations- 
systeme. Band 1: Hardware, Software und Daten. Wich- 
mann. Karlsruhe. 429 pages. 
Bil R., Fritsch, D., 1992. Grundlagen der Geo- 
Informationssysteme. Band 2 : Analysen, Anwendungen 
und neue Entwicklungen. Wichmann. Karlsruhe. ca. 429 
pages. 
Burrough, P.A., 1986. Principles of Geographic Information 
Systems for Land Resources Assessment. Oxford Science 
Publications. Monographs on Soil and Resources Survey No. 
12. 194 pages. 
740 
Reimer, U., 1991. Einführung in die Wissensreprásentation. 
B.G. Teubner. Stuttgart. 313 pages. 
Rich, E., 1988. KI-Einführung und Anwendungen. Mc Graw 
Hill. Hamburg. 467 pages. 
(turf