Ww o — Ww UV -
A^ —
^: 7. == UCL WU AO
WV e «J
be NA VE. NEL. N^ WM, NM X
— Ud SK We
On the other hand, the advantages of the digital approach
will increase. The way to get the city structures is not
uniform and standard. Finally a semi-automated way was
chosen (Guretzki, Erhardt, 1996). Hereby the height
determination and the horizontal determination of the
buildings is separated into different steps. Firstly two very
dense height models will be created by matching
procedures, the digital terrain model (DTM) without
man-made or natural features and the digital elevation
model (DEM) with all the features above ground. Secondly
the building outlines will be digitized on-screen by using
orthoimages (which are created with the help of the DTM).
A special technique enables avoiding the distortions due to
projection and height differences. This is done by linking
the orthoimage window and the DEM window, so that the
operator can easily move the outline to the correct place.
Special software combines now the elevation model and
the building polygons. By using statistical reports the
heights per building are defined, the location of
misclassified buildings (no representative height could be
defined) are shown and the heights are added as attributes
to the vector dataset. In addition the kind of roof (flat roof or
top ridge) and natural features greater than 3 m above
ground can be evaluated.
In the first phase of the planning process of microcells four
German cities (Berlin, Düsseldorf, Frankfurt and Munich)
with a total area of approximately 900 km? have been
evaluated with the two described methods. During 1996 we
expect the delivery of further city structures. This will be the
cities of Hamburg, Hannover, Dortmund, Cologne,
Dresden, Mannheim and Nuremberg with a size of
approximately 1.500 km? in total.
3.1. Experiences
As it was expected the analytical approach was very
precise and fulfils the requirements completely. Base of the
evaluation were aerial images in the scale range of
1:15000 up to 1:18000 (recorded with a wide angle
camera). The controlling of the data produces some less
errors or misclassifications.
The digital approach was carried out with aerial images in
the scale range of 1:12500 up to 1:15000 (recorded with a
normal angle camera). The images have been scanned
with about 25 um pixelsize, that led to a pixelsize at the
ground of about 30 cm. The results are a little less precise
in comparison to the analytical methods, but the
requirements are completely fulfilled. Nevertheless
automatic procedures require a greater expenditure of
controlling in order to ensure reliable data.
The greater scale and the greater focal length in
combination with the manual digitizing led to no significant
reduction of generation time and costs up to now. However,
the digital approach is continuously improving. The
progress in automatic triangulation and improvements in
the field of image recognition (at least the automatic
detection of simple buildings) will increase the advantages
of the digital method.
There are still some general questions concerning the
propagation model that must be solved. The 2.5D solution
used up to now causes some problems with tunnels,
buildings on pillars etc. Another question is the general
type of data. It was found out that the zigzag building
outlines caused by the nature of raster data lead to
801
additional diffraction edges that do not exist in reality.
Currently, work is going on to examine whether vector data
will show more realistic behaviour of the model. To be
prepared for this cause the city structure data was also
ordered as vector data. Together with the height
information, it would be easily possible to generate real 3D
data.
Over the last years the market of the telecommunication
industry has seen an almost exponential increase. This has
brought also a great benefit to both the private and the
public sector of geographic data supplement. The planning
and modelling of radio networks is just possible and
economic by using modelling software and digital
(geographic) data. Terrain models, in urban areas in
combination with city structure data help to increase the
correlation between predicted and measured signal
coverage. So the diffraction and reflection effects on radio
waves caused by various terrain and building heights can
be modelled. The lack of existing city structure data leads
to great efforts on the mobile telecommunication sector to
generate this data. The public administration with its maxim
to generate high precise (and of necessity expensive) data
is of lesser help. For Mannesmann Mobilfunk it is more
important to get data in volume and in acceptable time.
Moreover it is found out, that often available public digital
data sets are so expensive that a new generation (with the
advantage of defining the own appropriate requirements)
will be cheaper.
City structure data is not only useful for planning radio
coverage. It can also be used in the wide field of pollution
investigations, environmental planning, transportation
services, marketing and so on.
References
Baudoin, A., 1995, The SPOT programme: today and
beyond 2000, In: Fritsch/Hobbie Photogrammetric Week
‘95, pp. 63-74
Cichon, D.J.,Kürner, Th.,Wiesbeck, W., Modellierung der
Wellenausbreitung in urbanem Gelànde, In: Frequenz, vol.
47, no.1-2,pp. 2-11, January 1993
Dang, T., Jamet, O., Maitre, H., 1994, Using disparity
models and object models to improve stereo reconstruction
of buildings, In: ZPF 5/94, pp. 167-173
Feistel, M., Baier, A., 1995, Performanceofa
three-dimensional propagation model in urban
environments, Sixth IEEE International Symposium on
Personal, Indoor and Mobile Radio Communications 1995,
Vol. 2, pp. 402-407
Fritz, L., 1995, Recent developments for optical earth
observation in the United States, In: Fritsch/Hobbie
Photogrammetric Week '95, pp. 75-83
Guretzki, M., Erhardt, H., 1996, Erfassung städtischer
Gebäudehôhenmodelle mit Einsatz der digitalen
Photogrammetrie, Bildverarbeitung und ARC/INFO, ESRI
User Conference Germany
Haala, N., 1995, 3D building reconstruction using linear
edge segments, In: Fritsch/Hobbie Photogrammetric Week
'95, pp. 19-28
Huertas, A.,Nevatia,R., 1988, Detecting buildings in aerial
images, C.V.G.I.P., 42, pp. 131-152
Lang, F., Schickler, W., 1993, Semiautomatische
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996