dancy. Out of different considerations the range measurement 
precision of the Sercel receiver in dynamic application can be 
assumed reasonably to be co= 0.006 m. This value is twice as 
large as the precision obtained by a stationary receiver and 
depends mainly on the velocity of the aircraft. With this as- 
sumption the accuracies of the coordinates given in table 2 are 
calculated (Frieß 1988). 
  
Table 2: Estimated internal accuracy co» [m] of GPS 
coordinates from least squares adjustment with 
assumed range measurement accuracy of coz0.006 m 
X Y Z 
rms of strips 3,4,5 0.022 0.009 0.017 
rms of strips 1,6,7 0.020 0.009 0.020 
  
  
  
Comparing the estimated internal accuracies from the ARI model 
in table 1 without any a priori stochastic information and from 
the least squares adjustment with apriori knowledge of the 
sigma naught (table 2), we note a very good agreement. It 
demonstrates that the ARI model is suited to be applied for 
modelling the dynamic characteristics of GPS coordinates during 
a photogrammetric flight and that the results are realistic. 
All subsequent evaluation of the sensor data, for example the 
aerotriangulation, is based on the filtered GPS coordinates. 
These filtered data with their stochastic informations are 
achieved from the ARI-model. Table 3 summarizes the estimated 
accuracies of the filtered coordinates and their autocorrela- 
tion coefficients. It is an important result that in this exam- 
ple only the correlations from one to the next point (within 
0.6 sec.) remain significant. 
  
Table 3: Accuracy criteria of filtered GPS data 
Estimated accuracies ox of filtered coordinates [m] 
X Y 2 
rms of strips 3,4,5 0.017 0.007 0.021 
rms of strips 1,6,7 0.013 0.006 0.016 
Correlation coefficients of filtered data d=0.6 sec 
X Y Z 
r (1d) 0.46 0.69 0.52 
r (2d) -0.06 0.11 720.01 
r(3d) 0.06 T9715 -0.10 
  
  
  
86