originals very often contain features of most varying
qualities. The planimetric representation of our basic map
at the scale 1:5,000 is an example of such a scanning
original. Here we find road signatures, building signatures,
water areas, divers land-use-areas, topographic and trigo
nometric single symbols, etc.
A full-automatic extraction of certain object categories
from a binary image meets with many difficulties and tests
which can be compared have not been very sucessfull so far.
Some available software or some proposals in published
papers do use the line width or the line length or the
node-characteristic of signatures for the classification
of the object-qua1ity. These techniques are sometimes
usefull for the classification of characters or single
cartographic symbols, but not for the extraction of building
or street signatures.
At the scale 1:5,000 many objects are represented with
their natural or simplified contours, not in form of a
single uniform symbol. Furthermore we have the problem that in
our basic map several topographic objects are contoured
with lines of the same line width. So the line width indi
cates no feature quality.
TESTS ON THE RECOGNITION OF AREA PERIPHERIES
AND AREA QUALITIES
At the time we are engaged in investigations of the practi
cability of a certain solution (Lichtner 1981) for the
planimetric representation. For this solution we need a
colour-scanner which are used in the reproduction and
printing industry and an image»processing system. Both are
available on the market and with these equipments it is
possible to get feature-se1ected representations or image
matrices of the planimetry map original.
In the following the procedure chosen for this operation is
shortly outlined (see fig.l). First the planimetric re
presentation is copied onto a medium which can be drawn on
folt-tip-pen or another pen of special making. Then in this
copy of the planimetry all areas enclosed by peripheries
are manually encoded with a colour dot line or point
related to the quality of the represented object. For the
road net usually only one single colour dot is sufficient.
Building signatures - because of their hachures are made up
of severals single areas enclosed by peripheries - are not
encoded with colour dots. That's the class of uncoloured
areas. The manual colour-encoding of all these small areas
would be very time-consuming and susceptible to mistakes.
The vacant part of a real estate with a building is encoded
with a colour-dot.