ngly 
etric 
ation 
e all 
von 
der 
und 
otos 
ner 
otos 
rain 
198. 
rom 
the 
ms 
TM 
not 
be 
per 
jure 
ach 
the 
ing 
  
Bridge 
NE vile INR 
Aerial Photo | Aerial Photo 
Correct Orthophoto 
rdouble mapping 
displacenent-y Conventional Orthophoto {| Conventional Orthophoto 
  
  
  
  
  
Correct Orthophoto 
  
  
  
  
  
  
  
  
  
a) Bridge k> Bullding 
Fig.1: Displacement and double mapping 
Therefore city orthophotos require measurement of all 3D 
man made objects stored in a digital building model 
(DBM). 
1.2 Related Works 
Geometric aspects of 3D-objects are intensively 
discussed in (Turner, 1992). Experiments with an object 
oriented model are presented by (Heytens, Sacchi, 
1993). (Li, 1994) gives an overview of several common 
3D-data structures. (Lang et al, 1995; Haala 1995) deals 
with problems of automated building extraction from 
aerial images. An approach to mapping of buildings is 
given in (Gruen et al, 1994). A technique of orthophoto 
generation using a DTM with breaklines is discussed in 
(Ecker, 1991). Visualisation techniques of buildings 
models are presented in (Kuhn, 1990) and rigorous 3D 
modelling of city landscape is done by (Gruber et al , 
1995; Ranzinger, Gleixner, 1995; Sinning, Gruen, 1995). 
2. PROPOSED SOLUTION 
The proposed solution for accurately mapped 
orthophotos requires the following processes: 
digital building model generation 
building orthophoto computation 
terrain orthophoto computation 
raster algebra 
The digital building model determines the shape of each 
building and is required for the subsequent process of 
building orthophoto generation. 
Since conventional orthophotos systems can hardly 
model building surfaces and do not care about hidden 
surfaces, an algorithm has been developed for this 
purpose. It will be presented in detail in chapter 2.2. The 
result of this process is an orthophoto of all buildings and 
a building mask indicating buildings in the aerial image. 
For the subsequent generation of the terrain orthophoto it 
is sufficient to apply the conventional orthophoto method. 
But the input is not the original aerial image but an aerial 
image with grey value 0 in building areas. This modified 
aerial image can be generated using the building mask. 
By using raster algebra as described in chapter 2.4 it is 
possible to combine terrain and building orthophoto. 
2.1 Digital Building Model (DBM) 
There are many types of 3D-objects visible in aerial 
images. Natural objects such as trees will not be 
considered in this work. An overview of man made 
objects (buildings) and their representations is given in 
figure 2. Similar homogeneous objects can be stored 
economically using implied topology. Complex objects 
might be approximated by mathematical functions and 
construction rules. 
  
a) b) 
Fig.2: a) Complex object 
b) Mass parametrical objects 
For large scale orthophotos objects of figure 2b are 
important. A flexible solution to model this types of 
objects is boundary representation (B-Rep). This 
technique uses geometric primitives to describe objects 
boundaries. It is sufficient to choose triangles and 
quadrangles as geometric primitives. Beside from 
geometric shape the building model manages also 
attribute information for each geometric primitive. The 
digital building model distinguishes between primitives 
that are visible in the orthophoto (e.g. roof) and invisible 
(e.g. wall, vertical roof). Figure 3 shows a building and its 
representation in the digital building model. 
  
  
> 
NT. © 
s 
> 
E 5 
fae 
Coordinates Wall Roof 
X, M noi: 1342, 71:39, 
XV. Zona V4» 3564, 39105, 
Xu din XYZ, 5786. 5107 
XYZ: "XYZ, 7.128 71091 
X, Ys 25; Xo Yio £10 
  
  
  
  
  
  
  
Fig.3: Representation of a building 
in the digital building model 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996