Assuming that the image coordinate measurements are independent 
and have the same accuracy (standard error) the weight- and correlation 
numbers of the expressions (99)— (102) can be obtained according to 
their definitions. 
The weight numbers are found as the square sums of the coefficients 
of the measured image coordinates 
Quum (103) 
Quy. = 17 (104) 
5 
Qu = 57 (105) 
Qus (106) 
The correlation numbers are the product sums of the coefficients of 
corresponding image coordinates. All correlation numbers are zero. 
Above the error transformation between the individual pairs of photo- 
graphs was treated under the condition that only two transfer points 
were used. If more transfer points are used, attention has to be paid 
to the problem how the discrepancies within the points must be treated. 
Each point in addition to the two necessary points means two redundant 
observations, one in z and one in y. As usual, the discrepancies are 
most conveniently treated in accordance with the method of the least 
squares. This means that such values of the elements of transformation 
b 
the residual discrepancies in the transfer points a minimum. The problem 
dx, dy, dx and —— must be determined that make the square sum of 
requires that normal equations are formed and solved in order to express 
the corrections of the elements of transformation as direct functions 
of the coordinate discrepancies in the transfer points. General solutions 
of such normal equations and complete determination of the error 
propagation have been performed by HALLERT 1944. If the transfer 
points are located symmetrically with respect to the center point of 
the small strip in which the transfer points are located (the nadir point 
of image % in the image sequence £ — 1,%,% + 1) the solution of the normal 
equations becomes very simple. In this way an arbitrary number of 
transfer points can be treated very conveniently in accordance with 
the method of the least squares. The discrepancies in the transfer 
points then have to be expressed in terms of the errors of the original 
image coordinate measurements in order to investigate the error propaga- 
  
  
  
N