jon 
amas the 
(Xf major 
nishes to 
le frame 
tripod is 
following 
ibuted in 
era to be 
servation 
sequently 
following 
1 by this 
the single 
1 with a 
g. 0). Tie 
yrientation 
tion). The 
siduals of 
ie stability 
system is 
ismounted 
  
4. ORIENTATION OF PANORAMAS 
4.1 Definitions 
The orientation of panoramas is decsribed in analogy to 
standard methods of close-range photogrammetry based on a 
cartesian coordinate system. Fig. 10 illustrates the orientation 
parameters of a single panorama. The essential parameters are 
defined by the position (X,,Y,,Z,) of the cylinder projection 
centre, and the initial direction q of the panorama image. The 
tilt angles & and « describe a possible deviation of the tripod 
rotation axis from the vertical direction. Using bubbles these 
angles should be small. 
In order to get the orientation parameters of several panoramas 
simultaneously, tie points between these panoramas are 
required. The orientation accuracy mainly depends on the use 
of natural or signalised points. As an example, Fig. 11 shows 
measured signalised tie points. 
A 
d 
  
  
  
  
  
  
  
   
  
  
  
  
  
    
  
uw 0202. 
Fig. 11: Measured tie points in a panorama 
Tie points represent angular directions within a panorama. For 
the adjustment process the observation equations (6) are set up 
according to Lisowski & Wiedemann (1999). Due to the pre- 
adjustment of the camera/tripod unit it is not necessary to 
consider the excentricity between rotation axis and perspective 
centre. 
Additional required information are, on the one hand, an 
absoulte dimension of the coordinate system. This can be 
derived by one or more measured object distances between tie 
points. On the other hand, the datum of the coordinate system 
has to be defined. This problem is solved by fixing the 
orientation parameters of one of the panoramas, e.g. by setting 
all values to zero. 
In this way a "model coordinate system" with real dimensions 
is achieved. If this system shall be transformed into a global 
coordinate system (e.g. building coordinate system) a 3-D 
transformation based on control points can be applied. 
Alternatively, additional constraint equations can be introduced 
into the adjustment process. 
4.2 Orientation process 
The non-linear orientation process requires appropriate initial 
values of all unknowns. Usually an iterative calculation of 
approximate values is performed for photogrammetric bundles, 
such as a combination of relative orientations and resections. In 
our case there is no calculation of approximate values but, we 
assume that all panoramas cover the same room. Thus, all 
panoramas are assigned with equal coordinates of the cylinder 
origin and equal tilt angles (70) except the inital direction q. 
The first panorama has q-0, the subsequent panoramas are 
oriented with respect to the first one, whereby an approximate 
orientation angle of + 45° is sufficient. 
The initial values of unknown object coordinates are calculated 
from the first panorama with constant distance r, that equals, 
for simplicity, the average point distance in object space as 
shown in Fig. 12. The light grayed area represents the imaged 
room. 
  
  
Fig. 12: Grafical sketch of the start situation for initial 
values of orientation process 
  
  
Fig. 13: Grafical sketch of the iterative adjustment process 
  
  
  
  
  
Fig. 14: 3-D view of the orientation result 
Fig. 13 illustrates the progress of the orientation process (final 
result in Fig. 14). It is clearly visible that the final orientation 
—185—