HET measuring error of the second operator 
  
  
  
  
  
  
  
The mean error of the individual measurement and that of the mean were calculated out of the differen- 
ces between the double measurements. The subsequent computations were based upon this mean of two 
measurements. The mean elevations were fitted onto one another in different combinations, and we 
compared the discrepancies of both operator’s measurements as well as the discrepancies of the mea- 
surements performed by the same operator with diapositives and negatives. Therewith we distinguished 
between control points and border points. After these “free” transformations we fitted the elevations ob- 
ll served onto the given control points, separately for each stereoscopic pair. In this way we computed 
i the improved elevations of the border points, and subsequently compared again the fittings performed 
by different operators, as well as both the plottings performed by the same operator with diapositives 
and negatives. In most cases the border points revealed greater errors than the control points. As a 
remedy we finally transformed the measurements onto the control points in such a way, that moreover 
the square sum of the differences in the junction of successive models became also minimum, Thereupon 
Il the residual errors showed a better coincidence. 
All these errors consist of a series of individual errors. Beside the real measuring error m we distinguish 
the error of identification i (involving an error of orientation), the defect of the photograph (defect of 
|| Elm) f, the error pertaining to the control point p; (error in elevation), and the defect in copying k 
| occurring in the production of diapositives. The difference between two measurements of the same stereo- 
scopic model performed by two operators consists of the measuring errors m; and me pertaining to 
i both the operators, and the errors of identification i; and is. If we now consider the measurements of 
H the same stereoscopic model performed by the same operator, but in the first run from diapositives 
I and in the second run from negatives, the difference only consists of the measuring error and the error 
ll of identification pertaining to the first operator. But in this case the defect in copying must also be 
taken into account. On the other hand, if we consider differe nt stereoscopic models performed by 
the same operator, the difference between the junctions of successive models consists of the measuring 
errors mi or ma, of the error of identification i; or is, and of the defect of film f. With the dia- 
Ll positives the defect in copying k is connected therewith. If the stereoscopic models are fitted onto 
HE control points, the error of those control points has to be added. On the other hand, if we compare two 
it of such fittings performed by different operators, this error of control points and the defect of film are 
eliminated. There only remain the measuring error and the error of identification, involving the errors 
of orientation. By this method we have computed a total of 39 different combinations. An analysis of the 
results reveals that no defect in the process of copying exists, nor could we find an error of the control 
points which is worth mentioning. The values of the remaining mean errors are as follows: 
measuring error of the first operator m = 9cm 
mo — 10 cm 
error of identification pertaining to the first operator ij = 19cm 
error of identification pertaining to the second operator i2 = 14cm 
defect of film effecting a levelling error Ff = 27cm. 
The observations of both the operators were of equal quality. The great error of identification is rather 
striking. It involves the errors of orientation. Including the measuring error there appears a mean 
levelling error of 19 cm, which is slightly exceeding 0,1 per thousand of the flight altitude (hg = 1 800 m) 
The measurements with diapositives and negatives coincide with each other. Therefore we finally fit 
again all measurements one after the other onto the control points, but as already mentioned above, 
En by including the border junction of successive models into the minimum condition. We computed again 
the errors of the individual plottings out of the deviation from the appropriate mean values. The defect 
HE of film, effecting levelling error — converted into a linear unit and measured in the photograph — amounts 
| to about 11u. The material of photographs available, however, is by no means sufficient to draw 
definite conclusions. While computing we found certain major difference 
s which were rather striking 
and out of place. 
We should have an independent checking, if we should measure the defe 
the y-parallaxes. For this purpose the formulae for the relative orientati 
which after adjustment with the parallaxes in sev 
cts of film directly as errors in 
on of stereopairs may be used, 
eral points will reveal the residual errors of these values. 
Considering the uncertainty of our results we may round off the errors and obtain the following values: 
6 
  
    
T^ m- 
fta t5 DD