Na 
  
nd the 
  
 dra- 
stones 
  
merged to obtain meaningful areas (eg road areas). The 
merging can be constrained by relations of the topo- 
graphic objects (eg roads build up a network) that are 
denoted by the areas. 
6. OUTLOOK 
A model and its representation for the interpretation of 
large-scale maps and first results for the interpretation 
of a cadastral map were presented. 
One of our major goals is the determination of the 
topographic objects. Support can be gained from the 
following principles that describe relations between 
topographic objects. Together with the topographic ob- 
jects these principles can also be used for the analysis of 
aerial images. 
The first principle is based on a relation between the type 
of basic objects that can be regarded as areas and their 
functionality: Basic objects that together build up a "mo- 
saic of areas” (eg building, meadow, field) and basic 
objects that build up one or more "networks of long 
areas" (eg road, railway, waterbody). Based on this dis- 
tinction relations exist between the objects building up 
the network of long areas on the one hand and the types 
of networks (eg rivers naturally build tree-like networks, 
roads build networks with a hierarchy according to the 
road types) and the types of the crossings (eg rivers cross 
railways and roads only by overpasses or underpasses) 
on the other hand. 
The second principle is that there exist relations between 
parts of the network of long areas and some kinds of 
objects (eg parking spaces are connected with the road 
network). 
The third principle deals with objects in the network of 
long areas. These objects are often accompanied by 
different parallel objects (eg traffic lanes and pavement 
track). 
The fourth principle is concerned with objects from the 
image mosaic. Their forms are related to their type 
(areas like buildings have well defined forms). The 
orientation of an object is related to the orientation of 
other objects (eg a house is often oriented parallel to a 
road). 
The fifth principle is a physical exclusion relation of 
some kinds of objects on some objects that can be regar- 
ded as areas (eg a pylon is not allowed to be placed on a 
highway). 
585 
7. REFERENCES 
Antoine, D. (1991): CIPLAN: A Model-Based System 
with Original Features for Understanding French 
Plats. Proc., ICDAR '91, First Intern. Conf. on Docu- 
ment Analysis and Recognition, Saint Malo, pp. 647- 
655. 
Domogalla, U. (1984): Ein Expertensystem für die auto- 
matische Erfassung von technischer Graphik. Proc. 
DAGM/OEAGM, pp. 297-303. 
Ilg, M. (1990): Knowledge-Based Interpretation of 
Road Maps. Proc. 4th Int'l Symposium on Spatial 
Data Handling, Zürich, pp. 25-34. 
Illert, A. (1991): Automatic Digitization of Large Scale 
Maps. 1991 ACSM-ASPRS, Vol. 6, Auto Carto 10, 
Baltimore, pp.113-122. 
Joseph, S.H. (1989): Processing of Engineering Line 
Drawings for Automatic Input to CAD. Pattern Re- 
cognition, Vol. 22, No. 1, pp. 1-11. 
Kilpeláinen, T. (1988): Automatic Objects Recognition 
in Line Maps. Int. Arch Photogr. RS, Part B10/IV, 
pp. 327-336. 
Maderlechner, G. Jeppson, O. (1988): Representation, 
Classification and Modelling of Graphs for Efficient 
Pattern Recognition in Line Images. 9'th Int. Conf. 
on Pattern Recognition, Rome, pp. 678-680. 
Mulder, J.A. (1988): Discrimination Vision. CVGIP 43, 
pp. 313-336. 
Niemann, H., Sagerer, G.F., Schróder, S., Kummert, F. 
(1990): ERNEST: A Semantic Network System for 
Pattern Understanding. IEEE T-PAMI, Vol. 12, No. 
9, pp. 883-905. 
Schmitz, W. (1991): Automation der Katasterkarten 
mittels Scanner und Mustererkennung. In: Geo-In- 
formatik, Schilcher (ed.), Siemens, pp. 181-186. 
Smith , R.W. (1987): Computer Processing of Line Im- 
ages: A Survey. Pattern Recognition, Vol. 20, No. 1, 
pp. 7-15. 
Suzuki, S., Yamada, T. (1990): MARIS: Map Recog- 
nition Input System. Pattern Recognition, Vol 23, 
No. 8, pp 919-933.