D. M 5 ir à Als 2X PR ang AB 
Fig.2: city of Munich 1x1m* building dataset 
Furthermore, an already available data set (from the 
Institut für Rundfunktechnik München) based on a city map 
was tested. This data was generalized in such a way, that 
only the outlines of building blocks and the average 
number of floors per building was determined. By assuming 
a typical height for a single floor, the relative height above 
ground could be derived. The combination with a terrain 
model led to absolute heights. However the accuracy was 
the poorest of the tested four datasets. Figure 3 shows the 
city map dataset. Although the plots are black/white, the 
different accuracies and the degree of specification of the 
two displayed datasets is recognizable. 
  
   
2.4 Results 
The three-dimensional propagation model used for 
fieldstrength prediction was fed with the different datasets 
(Feistel, Baier, 1995). For comparison of the results some 
drive tests have been carried out. Hereby the fieldstrength 
was measured by a receiver every 10 ms along the 
measurement route (driving route) and the geographical 
coordinates were delivered by the Travel Pilot navigation 
system. 
It was shown that the comparable 10x10 m“ and city map 
datasets will not fulfil the requirements of precise network 
planning. The use of the 1x1 m? pixelsize caused a huge 
amount of data (e.g. a ctiy size of 200 km? led to about 800 
MByte uncompressed data) and large computing times for 
propagation. Moreover the 1x1 m? data is not automatically 
better than the 5x5 m? data in the sense of propagation. 
The planning of radio cellular networks is a complex 
business with the input of many different parameters and 
their mutual dependence. So the prediction accuracy will 
not increase in line with the accuracy of the geographic 
data. Therefore the 5x5 m? dataset was the best 
compromise of both economical and accuracy/storage 
aspects. 
In total the following requirements are fixed (in respect to 
the model, which is now used): 
- combined dataset of both terrain and building heights 
- pixelsize of 5x5 m^ 
- horizontal accuracy of about half pizelsize, vertical 
accuracy + 2 m 
- generation of all buildings with a larger size than 50 m* 
and above ground height of more than 3 m 
- generation of the buildings as boxes with flat roofs 
(highest representative point) 
- perpendicular rise of heights between ground and 
building (height discontinuity) 
- divide of building blocks into several parts, if the height 
differs by more than 3 m 
3. Orders 
Main aim of the studies was to find out the necessary 
geographic parameters for planning microcells, not the 
method to fulfil these requirements. After defining these 
parameters some companies have been requested to 
propose appropriate generation methods and to approve 
their methods by demonstration data. Finally it was 
decided to select both analytical and digital photogrammetric 
methods for the first German cities to evaluate. 
Photogrammetric methods seemed to fulfil the requirements 
pointed out in part 2.2 at best. In addition the scanned 
aerial images can be an additional help for the planners. 
Resampled as orthoimages appropriate locations for base 
stations can be preselected in a simple way. The controlling 
of the delivered data is possible within a digital (stereo-) 
workstation with the same base from which the data was 
generated. Updating is possible by evaluating specified 
areas with up-to-date images in the future. 
The analytical way is unquestionably a precise, but also a 
relatively expensive method. The human operator with his 
experience and interpretation ability can hardly be beaten 
by automatic procedures. 
800 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
  
—^^—CI OD y) 0 rr c oe c — £r» 
Qt Tren TT > O 
— — 
SS: <r