) steps, 
so 
ie robot 
| system 
attached 
and is 
e whole 
| record 
rdinates 
l. Every 
aly put 
points. 
| bundle 
tes as 
| world 
inknowns 
robot 
roximate 
t the 
stem and 
1 in the 
he world 
est fit 
1e robot 
process 
vith the 
covering 
t stops, 
2 image 
tar are 
e points 
tion in 
points 
ar point 
and the 
ed with 
in the 
between 
ve the 
hat the 
| in the 
still in 
ts show 
]. Major 
of the 
cameras 
, change 
loss in 
is not 
4. OPTICAL CORNER FITTING MEASUREMENT SYSTEM FOR 
AIRCARFTS 
4.1 The Task 
A corner fitting of an aircraft can be described 
as an angle iron that connects wing and body. Due 
to the high sensitivity of this junction the 
corner fittings have to be manufactured very 
precise. 
Because of mechanical limits every aircraft of 
the same type is slightly different to each other 
so the shape of the corner fittings has to be 
adapted to every aircraft. Therefore a 
measurement system is necessary to measure the 
spatial position of the surfaces, the corner 
fittings are attached to. With the result of this 
measurement every corner fitting. can. be 
manufactered unique with a numeric controlled 
milling machine (Rüther, 1989). 
The major requirements to this system are an easy 
handling, high accuracy and high recording and 
processing speed. The problem is a very small 
work space in the wing of the aircraft. The 
corner fittings have a size from 0.5m x 0.3 m to 
0.8 m x 0.5 m. The whole work space is about 3.0 
m x 1.0m x 0.8 m. In this workspace the user has 
to fulfill the whole measurement work. 
4,2 System Configuration 
Due to the very small work space the image 
recording device has to be very small and the 
processing unit has to be placed outside the 
aircraft. The camera module is a JenaCam 28 CCD 
camera with shutter and an external pixel 
synchronous control unit and 8.0 mm lens. 
The processing unit is a PC based transputer 
network with four transputers for parallel image 
processing. 
The surfaces are targeted with a film. A special 
pattern of coded targets is printed on the film. 
Every target consists of a circle in its center, 
the actual target and a coded ring around this 
circle that keeps the coded identification 
information. The targets are set up in a grid of 
about 20 mm grid width. So the surfaces of every 
corner fitting are signalized with about 150 to 
200 targets. 
Additionally a scale bar has to be put in the 
scene to allow a scaling of the derived point 
coordinates. 
4.3 Measurement Procedure 
The measurement procedure starts with attaching 
the target film onto the surfaces to be measured. 
Then the user captures one image after the other 
of the surfaces. Every image is processed 
automatically after it is recorded. The 
transputer system allows a parallel recording and 
processing without any pauses. The communication 
with the user working in the wing and the 
processing unit outside is performed by a small 
monitor that shows a life image of the camera and 
additional information about the actual 
processing status. 
  
   
   
   
  
    
    
     
     
    
     
    
     
     
   
    
   
     
    
    
     
    
    
    
    
     
        
     
     
     
     
   
     
   
  
    
    
   
   
     
     
    
    
   
   
    
     
  
  
       
      
      
       
   
  
  
The image processing runs automatically. In this 
process the coded targets in the images were 
first detected, then the coordinates of the 
center point measured and finally the code ring 
decoded to get the identification information for 
every target. 
9, 
Fig.4: Coded target 
If all necessary images are recorded, the image 
coordinates are also available. Before the 
coordinates of the points can be calculated using 
a bundle adjustment, initial values for the 
object coodinates and the exterior orientation 
have to be provided. 
The initial values for the coordinates are 
already available because the target film is 
attached always at the same position of the 
aircraft. Because of the same shape, coordinates 
once measured can be used as initial values every 
time. The initial values for the exterior 
orientation can be derived by a spatial resection 
over the initial coordinates. 
The following bundle adjustment delivers the 
coordinates of the surface points which can be 
used for the numeric controlled milling machine. 
5. CONCLUSION 
These examples have shown the wide range of 
possible applications of optical 3-D measurement 
systems. More systems are already in the planning 
stage and with a growing number of installed 
systems optical measurement techniques will 
become a significant tool in quality control and 
inspection. 
6. REFERENCES 
Bósemann,  W., et.al., . 1990. Photogrammetric 
investigation of CCD cameras. IAPRS Zurich, Vol. 
28, Part 5/2, pp. 816-822. 
Peipe, J., 1989. Measurement of Kinematic Robot 
Performance by a Hybrid Photogrammetric System. 
Congress on Optical 3-D Measurement Techniques. 
Vienna 1989. 
Rüther, R. Wildschek, Re, 1989. Digital 
Photogrammetry for the Acquisition of Surface 
Data for Automatic Milling Processes . Congress 
on Optical 3-D Measurement Techniques. Vienna 
1989. 
Schneider, C.-Th., Sinnreich, K., 1990. Concept 
of an Optical Coordinate Measurement Machine. 
IAPRS Zurich, Vol. 28, Part 5/2, pp. 816-822. 
Wester-Ebbinghaus, W., 1985. Bündeltriangulation 
mit gemeinsamer Ausgleichung photogrammetrischer 
und geodàtischer Beobachtungen. ZfV, 110, pp. 101 
- 111. 
MOS PME time 
E