International Archives of the Photogrammetry, Remote Sensing and Spatial Information 
  
Sciences, Vol XXXV, Part BS. Istanbul 2004 
  
6.3 Sensitivity of deformation detection 
Set 20, 12 and Set 8 were examined to check whether or not 
0.050 mm displacement could be detected. Table 2 shows the 
measured movement of the three points, i.e. Points 101, 102 and 
103, in Block B obtained in the test to check whether the three 
points moved for Sets 20 00 and 20 50. Because there are no 
absolute datum points, the scale of these figures is approximate. 
The three points include not only parallel displacement but also 
rotational displacement. 
F value of Equation (12) was obtained to check whether or not 
the three Points 101, 102 and 103 had moved. The results are 
shown in Table 3. F was calculated for the case where only 
Point 101 was tested and for the cases where two points and 
three points were tested. Testing capability decreased due to the 
rotational movement of the three points. Table 3 shows F value 
Ta (95%) at X = 5%, square root of non-centrality 5,, and 
square root of non-centrality ó that gives 8 = 80% at Œ = 5%. 
The following are found from Table 3. 
* Sensitivity decreased when three points were used, because 
there were rotational components. It is better in practice use 
to avoid assuming rigid body displacement of multiple points. 
Thus, testing should be conducted for each point.. 
* If the testing capability of 4 = 20% is required, observations 
of three points are necessary for Set 12. 
« Even if testing capability is reduced, the same number of 
observations as Set 12 is necessary for testing each point. 
« When three points are usable, there is sufficient sensitivity for 
Set 8. 
6.4 Detection of gross errors 
Results of F testing by the data snooping for Set 20 00 are 
shown in Figure 3. One gross error is detected. An image of the 
gross error is shown in Figure 4. The point where @ = 5% for 
F(1,m — r, —1) had about Ta (5%) = 4, so the T’ value 
of the detected gross error is sufficiently large. 
The distribution of the square root of non-centrality is 
investigated for Set 20 00, Set 12 00 and Set 8 00, assuming 
that a standard deviation of image coordinates three times as 
large as a priori value with unit weight, 0.0005 x3 = 0.0015mm 
is provided to discrepancies. Figure 5 shows an example of Set 
12 00. Although the non-centrality slightly increases at both 
ends of the model, the mean, minimum and maximum figures 
are shown in Table 4. The effect of variations of camera 
configuration to the non-centrality is slow to react. A square 
root of non-centrality ó? for B=20% at the point where Œ = 
5%, is about 2.8. This is less than all values of every Set, so a 
sufficiently reliability is confirmed in any three sets. 
Table 4. The square root of non-central values ( Ó ) 
for three exposure configurations 
  
  
  
  
Set Mean Minimum Maximum 
Set 20 00 3.27 3.12 3.66 
Set 12:00 3.36 3.17 3.78 
Set 8 00 3.50 3.25 3.96 
  
  
  
  
  
  
80 T T 
  
  
  
  
Mumber of Target Images 
Figure 3. Results of F testing for Set 20 00 
  
Figure 4. Image of target with gross error (Center 
Square Root of Moncentrality 
Figure 
of target is slightly deviated) 
  
  
  
  
  
A i L J i À i L 
39 100 200 300 400 500 600 700 800 900 
Number of Target Images 
5. Distribution of square root of non-centrality; 
Distribution of non-centrality Ó when the square 
root of pre-variance of error sigmax 3= 0.0015mm 
is set to every observed values for Set 12 00. 
   
Internat 
This pa 
of slope 
Assumi 
applied 
interval 
mm di 
investig 
taken a 
requirer 
displace 
Fraser,C 
Topogre 
pp.1115 
Fraser, 
Photogr 
Mounta 
Kiameh 
NCC Ge 
Benzao, 
Networl 
pp.33-2 
Koch, I 
Linear | 
Koch, F 
Linear | 
Hattori, 
Procedu 
Photogr 
Vol.68,1