rithm 
is H) 
nask, 
mage 
pixel 
iginal 
range 
r the 
r this 
n be 
orrain 
aster 
y the 
nage 
also 
The 
o the 
nage 
have 
input 
to. 
  
face 
Iding 
ging 
os in 
only 
iden 
final 
| by 
rent 
lired 
  
For pixels mapped in both orthophotos an appropriate 
rule (e.g. average) has to be applied to compute the final 
grey value g, ,. 
3. ORTHOPHOTO PRODUCTION ENVIRONMENT 
3.1 Data Acquisition 
Currently data capturing for building models is usually 
done manually on analytical or digital photogrammetric 
stereo digitizers. ^ Approaches for automated data 
acquisition applying digital techniques are still under 
development (Lang, 1995; Haala 1995). Therefore 
simple procedures for manual acquisition on a 
stereoplotter integrated into a CAD environment are of 
great importance. Capturing of DTM-data and 3D man- 
made object data will be done simultaneously. 
The usage of existing data generally captured for other 
purposes (e.g. maps etc) is very difficult. In most cases 
these data are stored in two dimensional GIS. Therefore 
additional information such as 3D topology, elevation, 
etc. is required. Inconsistency of these data sets is 
another big problem. It is caused by data acquisition 
using different methodes, at different epochs for different 
purposes. 
Since data acquisition for building models is very time 
consuming, municipalities of big cities should set up data 
sets for multipurpose future usage. 
3.2 Modelling 
Fortunately many buildings (especially in suburban 
areas) are of simple shapes. In this case it is sufficient to 
digitise roofs only. Walls can be computed from roofs 
(eaves) and DTM (see figure 7) on the assumption that 
eaves and terrain are boundaries et walls. This methode 
is not to apply for overhangs. 
Three dimensional triangulation algorithms with 
geometric constraints (Halmer et al, 1996) are necessary 
for automatic generation of geometric primitives as 
described in chapter 2.2 (see figure 8). This method is 
used to model details within roofs. A test detects nearly 
vertical roofs to be classified as wall. 
  
Fig 7: House modelling by roof and terrain data 
  
e <P 
a) input b) result 
Fig. 8: Triangulation with geometric constraints 
  
  
  
  
  
For complex bodies, modelling must be done 
simultaneously with capturing. Therefore the integration 
of modelling and capturing modules is necessary. 
In some countries buildings are very similar and therefore 
it is possible to prepare standard buildings with same 
topology (Ranzinger, Gleixner, 1995). (Lang et al, 1995) 
proposes a method for semi automatic topology building. 
Digital photogrammetric methods such as edges 
extracting or matching might be extended to 3D-topology 
matching. 
3.3 Orthophoto Management System 
For citywide orthophoto generation (with several hundred 
thousand buildings and also several thousand digital 
photos and associated control data), a database system 
is necessary to manage all these data. 
This orthophoto management system should handle all 
relevant data: 
e aerial images including image properties (orientation, 
etc). 
e digital terrain model 
e digital building model 
Required data for orthophoto computation will be made 
available by the system. Another important task is data 
exchange (import, export) with other systems. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
[73^ (773 
s * 
ov 
: [5i 
LLL 4 LIII 
u 
— — 
> 
fe 
5 
Owl 
£ 
oc 
$ 
2 
2 P-L 
n| à jo 
LA] n 
^ Remote-Sensing-s t^. 
Be FA 
m A ru 
HF) 
- P-2 
no 
  
  
  
Fig.9: Object partitioning 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996