o € 
Recalibrating the taking camera after photography yielded unknowns Ac, A XH and A YH of the 
magnitudes shown in Table II. The standard errors of the unknowns were always in the order of 
+ 0.1 mm indicating a weak solution, 
  
  
Table II 
Ac/mm/ Ax /mm/ A yy /mm/ 
SINAR - 0,6 to + 0.3 - 0.2 to + 0,2 - 0,1 to + 0,1 
VEROPLAST -0.2 to - 0,2 0.0 0.0 
  
  
  
  
  
  
Plate Coordinate Measurement and Computation 
  
The measurement of the image points in the ZEISS stereocomparator PSK did put very 
high demands on concentration and the ability to abstract, The operator had to interprete the center 
of gravity of forces within a lump of solder or in the skew-whiff crossings of several wires, In 
addition to the measurements, sketches of the measured points and their neighbourhood had to be 
made up for later reference, Especially the definition of the system points was somewhat arbitra- 
ry due to their physical realisation and shape in the steel wire models, 
Another difficulty was what may be termed ''stereoscopic deception'. Due to the limited - 
field of view and the regularity of the net pattern it was possible to fuse two non-homologous net b 
wires to a perfect stereoscopic image. The deception could not be noticed during measurement , 
but only after computation of the residual parallaxes. For this reason and for the detection of 
errors in numbering - all observations had to follow a rigorous numbering scheme which served 
as point indices in further computations -checks and plausibility controls were part of the computer 
program. They shall be briefly outlined. 
The program's subroutine for analytic relative orientation follows in principle the known 
method of Schut [3]. From 5 homologous points a direct solution for the unknowns vo", w", x", 
by" and bz' is obtained. These are used as approximate values for a least square adjustment of 
condition equations with unknowns. The error equations 
are solved satisfying the function 
F = v'Pv - 2k'(B'v + Ax + w) = Min. (2) 
Differentiation of (2) with respect to v and x and substitution into (1) yields the system 
of equations 
B'P-1B A k -w 
2 (3) 
A! O x O 
After inversion of the normal equations the correlates k and the unknowns x may be 
solved and the corrections to the observations may be calculated from 
v = P-iBk, (4) 
For pairs of near vertical aerial photographs the elements of the weight matrix B'P-1B 
are nearly constant and equal to 2c? (c » calibrated focal length). Since small weight differences J 4 
have no influence on the results the computation of this matrix may be omitted. In close-range 
analytical photogrammetry, however, the rigorous adjustment of the relative orientation is worth 
while, especially if the ratio object depth/object distance is large and oblique photography is used. 
nd i