RSC 40 (top) 
Repeated Measurements 
Differences from mean value 
  
  
  
  
  
  
  
  
  
  
  
  
lil dx max C dx min B dy max El dy min 
Repeated Measurements 
Differences from mean value 
RSC 41 (center) 
  
  
  
  
  
  
  
  
  
I 
SM RO 
SR LV 
lll dx max C dx min & dy max & dy min 
Repeated Measurements 
Differences from mean value 
RSC 42 (bottom) 
  
  
  
  
  
  
  
  
  
variations 
  
lll dx max Ni dx min B dy max Ei dy min 
Fig. 6. Repeated measurements: image coordinates 
NM: no sensor movement; SM: with sensor movement; RO: with 
rotary movement; SR: sensor and rotary movement; LV: light 
3.2 Distance Measurements 
In this part of the investigation the system accuracy for 
distance measurements shall be verified. For this 
purpose different distances (approximately 700mm and 
1400mm) have been positioned at various locations 
within the measuring volume, oriented in different spatial 
directions. The nominal distances were measured using 
a first-order coordinate measurement machine (CMM). 
The distance bars are made of aluminium and cylindrical, 
retro-reflective targets are used for measurement. 
The measurements are performed using a sub-pixel 
algorithm for the precise detection of ellipse-shaped 
targets [Zhou 1986]. The measured image coordinates 
are processed by an spatial intersection program which 
determines the 3D coordinates of the target. 
The distance points have been measured in different 
configurations: 
NR measurements with different cameras, without 
rotations 
WR measurements with one camera and various 
rotations 
CO measurements with combinations of cameras 
and rotations 
20bs | measurements with 2 images 
3obs measurements with 3 images 
> 3 obs measurements with more than 3 images 
The number of observations per point varies between 2 
and 9 images. Fig. 7 shows a few results from different 
points of view. The first three clusters show the 
dependency of accuracy on the number of observations. 
The last three clusters show the results of different 
configurations. In each cluster the mean deviation is 
displayed, as well as the minimum and maximum 
deviations. Deviations are computed from the calculated 
distance between measured points and the nominal 
value measured by the CMM. 
Distance Measurements 
Differences from nominal distance 
  
  
    
  
     
  
   
  
  
  
  
  
  
  
mm 
0.20 
0.15 -—B - 
0.05 HN E - 
0.00 IN ES Ed ] | WS l N. | 
NR WR 2 obs 3 obs 
   
MM ds_min LS mean Bl ds max 
Fig. 7. Distance measurements 
NR: different cameras, no rotation; WR: one camera and rotations; 
CO: camera/rotary combinations; 2 obs: two images per point; 3 
obs: three images per point; >3 obs: more than three images 
   
     
      
  
   
  
  
  
  
  
  
  
  
    
    
  
     
   
    
   
  
  
     
   
    
   
    
   
    
   
      
These res 
are alwa 
observatio 
minimum 
measurerr 
than 0.2m 
A similar it 
effect of 
intersectio 
combined 
reliable an 
intersectio 
sufficient 
images. 
deviations 
The inves 
accuracie: 
configurat 
accuracy 
minimum | 
The proto 
Measurinc 
RSC digit: 
table. The 
this config 
System c: 
non-calibr 
combined 
measuring 
calibration 
cameras 
calibrating 
of 1.5um, 
Repeated 
out in ord 
the RSC 
accuracy 
achievable 
table is 
accuracy t 
The meas 
final accu 
results fro 
applied 
accuracy 
obtained 
sufficient 
turntable 
observatic 
always be