810 
Thereby the structural and numerical classification methods can in a combined 
manner be applicated in recognition of contour lines. The. structural method 
controls the processing procedures and the numerical method makes the decision 
which line segment belongs to the class to be expected. 
Recognition of figures and points. The figures and points that are used to build 
up the height informations of the terrain have individually to be extracted and 
classified by numerical pattern recognition. 
Wortbuilding. The individually recognized figures and points can construct the 
numbers which represent contour values and spot heights. The numbers can be 
reconstrue ted by syntactical classification. During this process some errors of 
text recognition, e.g. confounding 6 with 9, can be corrected in hand of know 
ledge based context method. 
Automatic assigning the heights to contour lines. After the recognition of con 
tour lines it is evident to which class they belong, but not the exact heights. In 
German Basic Map Series 1:5000 only the contour lines at the 10 meters ver- 
tival interval (called principal contour lines) have the full value, the other lines 
(called between contour lines) show only the last figure of the decimal position. 
At first the recognized heights should be assigned to the contour lines. There 
after the morphology and the topography of the terrain can be constructed de 
pending on the neighbouring principal contour lines and their features by means 
of knowledge based analysis. The morphological and topographical features main 
ly consist of the following three different terrain models: valley, top and slope 
that contain all kinds of terrain forms e.g. knob, bowl, saddle etc. Every bet 
ween contour line has to be examined to which terrain model it is belonging and 
the height of it must be carried over from the principal contour lines. 
Recognition of special symbols. The special symbols can be recognized by the 
numerical pattern recognition methods. 
Knowledge based controll of recognition procedures. This system is developed 
and tested in German Basic Map Series 1: 5000 and 1:20000. The different 
kinds of maps have different legends. Therefore the different symbols and the 
vertival interval have been used. The parameters and the recognition procedures 
are variable on different kind of maps. It is neccesary to build up a knowledge 
based system to automatically control the recognition processes of different 
kinds of contour line maps. 
Dipl. — Ing. Jun Yang 
Scientific assistant 
Institut of Cartography 
University of Hannover 
Appelstrasse 9 A 
3000 HANNOVER 1 
Federal Republic of Germany 
Tel. 0511/7623465