photograph 
  
digital target 
image model 
  
geometrically 
rectified digital 
target image 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
comparator 
coordinate system 
  
Fig. 4: matching of a digital target image model with a digitized target image and 
transformation of a photogrammetric image point defined in the target image model 
square matching. 
Both digital images are of the same geometric shape, 
because the target image model is computed in a met- 
ric coordinate system and the digitized target image 
is geometrically rectified by the transformation into 
the comparator system. The way this transformation 
can be performed depends on the type of the compara- 
tor. If, for example, the comparator is equipped with a 
reseau, the digitized target image is transformed into 
the comparator system by using the imaged reseau 
crosses. After eliminating a possible rotation parame- 
ter between the photograph and the comparator sys- 
tem the geometric parameters of the image matching 
consist only of two shifts. 
The radiometric parameters of the matching have to 
describe a linear transformation between both digital 
images, because from the ray tracing procedure only 
relative density information can be derived, level and 
scale are not achieved. Therefore the target image 
model has to be adapted radiometrically to the target 
image. Dependent on the density characteristic of the 
photograph, in some cases disturbing pixels around 
the edge of the target image model have to be elimi- 
nated before the matching (see Meid, pp.48ff). 
Using the outcoming matching parameters between 
both digital images the photogrammetric image point 
can be transformed into the real, digitized target 
image. With the transformation parameters between 
the digitized target image and the comparator coordi- 
nate system the point can finally be transformed into 
the last-mentioned one (Fig. 4). 
If this is done for each point of the complete image 
block, one iteration for the determination of compara- 
tor coordinates is complete. 
3.3.Orientation with Improved Comparator 
Coordinates 
If the accuracy of the bundle adjustment can be in- 
creased by introducing the image coordinates derived 
from the new comparator coordinates, the efficiency 
of the method is proved. Further iterations of the 
whole process (see Fig. 3) have to be performed as 
long as the accuracy of the bundle adjustment in- 
creases. 
4.Test of the Method on the Basis of an Example 
In this section the results of an underwater photo- 
grammetry project are presented. An orientation of 
three underwater images on 28 object targets (see the 
arrangement in Fig. 5, and see one of the targets in 
Fig. 1 on the left side), fixed at a frame, was required. 
The used camera was embedded in a housing, di- 
rected to a plane-parallel housing window with a 
thickness of 2.5 cm. This caused a dramatic decreas- 
ing of the image sharpness from the center of the pho- 
tographs to the rims (see also Przybilla et. al. 1988). 
Fig. 1 on the rigth side shows one of the blurred tar- 
get images. 
To demonstrate the results of the orientation by 
bundle adjustment, the residuals of a visual measure- 
ment and of an object based determination are com- 
pared (Fig. 6). The best visual measurement of three 
different persons led to the residuals shown in Fig. 6 
  
   
   
   
  
  
  
   
  
  
   
   
  
   
  
  
  
  
  
   
  
  
  
   
   
  
  
   
   
  
  
   
   
   
  
   
  
   
   
   
   
   
  
    
on t 
the 
ima, 
was 
base 
The 
this 
mea