Ri AU. SOME iN ac a ER EDU a e e oec d 
errors do exclusively appear at images no. 1 resp..3 and at points no. 1 resp. D: 
Case a: One gross error: at the image coordinate x' of control. point no.l 
of. image 1 resp. 3: vx; = ome 30: um 
Case b: Two gross errors simultaneously at the image coordinates x', y' of 
control point no.l of image 1 resp. 3: vx; = Vi z - 30 um 
Case c: One gross error at the image coordinate x' of non-control point no.2 
of image l' resp. 3: VX, = = 30 um 
Case d: One gross error at the image coordinate y' of non-control point no.2 
of. image 1 resp..3: Vy5 z - 30 um 
With respect to case b we act as if only" one' gross error-exists, although two 
gross errors have been introduced. 
The following parameters are chosen resp. computed for testing: 
& = 0,000 (0,1%) 
u 
- 801) * 8, = 0,8 (80%) 
A #14 M 4.14 
Critical value for data-snooping (with e it 0,001) from t(l-o,, 9): C(w.) + 3. 00. 
Hence the minimal gross error which can be detected with the probability ^ is 
(with Pg = I): 
1j Aoi 207 
Oo = 
( = = 
i 9 (vivi (av;v; 
Thus we obtain for the different arrangements and gross error situations the 
y] 
  
(27) 
values of Table 3 for the minimal gross errors. 
Table 3: Minimal gross errors to be detected with the probability 6 058. 
  
  
  
  
  
  
Arrange- Case a Case b Case c Case d 
ments Vx4 (min) VX4 (min)| vy1 (min) VX, (min) Vy, (min) 
[um] Dm] | [um] [um] [um] 
A 28.3 28.3 | 27.2 - 32.0 
B 31.4 94.2 ^2 19 29.6 - 32.4 
C 29.6 29,6 logy 48.5 27.3 
D 25.5 25.5 | 25-7 32.4 25.9 
E 27.2 27,2 | 26.8 31.3 26.2 
  
  
  
From Table 3 we see that the minimal gross errors are grouped around 30 um, except 
the cases Ac, Bc, Cc. That's the reason why we have introduced for all cases a -d 
a gross error of -30 ym. 
After the computation of all error cases with all network arrangements we have got 
the results of Table ^ for the global test (HD: E(85) = 224 Because the signifi- 
cance level a for data-snooping was kept constantly, the significance level a for