n of sur- 
els. Con- 
es in the 
( surface 
  
NA ANA 
Nd on the 
enerate a 
ay be the 
king such 
the shape 
ting CAD 
consist of 
elements 
available, 
ie existing 
isurement 
nation be- 
)ordinates 
idjustment 
age orien- 
  
Point coordinates 
Computer Aided Design Model 
Digital Surface Model 
Digital Orthoimage 
  
  
  
  
  
  
  
  
  
Fig. 4: Typical flow of geometric data 
tations and object coordinates. The object points have 
been connected to closed polygons to build up a CAD 
model, consisting of surfaces. 
The next step is to define a reference plane for the de- 
sired orthoimage. Usually it will be parallel to one of the 
buildings facades. On this reference plane an orthoimage 
grid with the required resolution has to be defined. Each 
grid element has corresponding object coordinates. The 
  
Fig. 5: Digital Surface Model (scale 1:400) 
607 
relations between the pixel matrix of the orthoimage and 
the object coordinates are usually linear. 
A second matrix, equivalent to the requested orthoimage 
and with the same grid distance, will be generated to store 
the data of the DSM. All surface elements of the CAD 
model are orthogonally projected onto this reference 
plane. For each grid element the elevation above or below 
the projection plane has to be calculated and stored in the 
DSM matrix. Higher elevations replace lower elevations. 
The result is a matrix with the maximum elevations over 
the projection plane. Fig. 4 shows the geometric data flow. 
Fig. 5 shows a pictorial representation of the used Digital 
Surface Model. Closer points on the surface are displayed 
darker than points further away. Linear features, like the 
fence around the platforms on the towers, are not repre- 
sented in the DSM. 
To refine the DSM matching techniques may be used. This 
will be necessary in areas where surface elements, e.g. 
sculptures, can not be described by a traditional CAD 
model. Whereas image based matching may refine the 
surface between edges, feature based matching may 
improve the three-dimensional accuracy of detected 
edges. This technique was not yet used in this example 
project, but it is under development. 
3. DIGITAL ORTHOIMAGE 
Before starting the calculation of the digital orthoimage, 
the data of the exterior orientation should be transformed 
into a new coordinate system. The result is a rectification 
coordinate system, with the XY-plane parallel to the 
reference plane, and the scale of the object coordinate 
system. If the reference plane is parallel to a coordinate 
plane of the object coordinate system, this can be done 
by changing axes. This rectification system reduces 
calculations in the time consuming detection of occluded 
areas and accelerates the rectification process. 
In the past the calculation time for digital orthoimages has 
been a matter of concern (Mayr & Heipke 1988). For a long 
time, anchor techniques have been used to accelerate 
the calculations. This is no more necessary because of 
the improvements in computer performance. Consequent- 
ly a pixel by pixel process can be used. Each pixel of the 
orthoimage has to flow through the calculations cascade 
illustrated in Fig. 6. 
Starting at the row and column in the image matrix of the 
orthoimage the metric coordinates of the point on the de- 
fined reference plane can be calculated. The results are 
the X and Y coordinates in the rectification system. 
The elevation of the point can be extracted from the digital 
surface model and we get a full 3D coordinate triple. 
If the ray between the object point and the rectification 
coordinates of the projection center intersects the DSM, 
the point lies in an occluded area and no grey value can 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996