of all non-modelled errors of the image coordinates, ji. e. errors from 
the atmosphere (refraction), the camera (optical and mechanical), the 
film, the actual measurements (operator) and the instrument (Plani- 
comp). By comparison the o, = 4 yum of case c) has essentially the same 
error budget (including the setting errors of the operator at the con- 
trol points) plus the effects of the digital point transfer. Thus we 
conclude that the error effect of digital transfer on the image co- 
ordinates of natural tie points amounts to 2.9 um (or more precisely to 
3.1 um, if the setting accuracy (about 1 um) of signalized control 
points by the operator is considered). Included are the random errors 
of the digital point transfer by image correlation, as well as the non- 
modelled systematic errors of the operation and of calibration of the 
digital arrays. Looking briefly at the value c, = 3.9 yum of case d) one 
can say the digital transfer of natural points and of signalized points 
by digitaT area correlation is practically the same. The difference is 
statistically not significant. (This case is to be distinguished from 
the direct digital capture of signalized points by specific image pro- 
cessing techniques without point transfer, which is outside the scope 
of this test and was not attempted here.) 
Looking at the absolute coordinate accuracies of the adjusted blocks b) 
and c), as assessed from 107 check points, the ux values of 3.3 um 
and 3.7 um, respectively, compare very well. Thos& Values are the re- 
sult of c, and the actual overlap and control configuration. They still 
include the errors of the geodetic coordinates (9 mm in the terrain, 
equivalent to 2.2 ym in the photoscale). The additional contribution of 
digital point transfer to the absolute accuracy of the block in this 
case amounts to 1.7 um. The apparent difference in absolute accuracy 
between the blocks c) and d) is statistically not significant, on the 
hypothesis that both have the same co. Finally it may be recalled that 
2.5 um in the photographs are equivalent to 1.0 cm in the terrain. 
Thus, all results are certainly in the range of high precision aerial 
triangulation. 
3. TEST BLOCK APPENWEIER 
3.1 The second test for digital point transfer utilized photography of 
the test block Appenweier from 1973, /6/. The wide-angle photographs 
(Zeiss RMK 15/23) have photo scale 1:7 800. Black and white film dia- 
positives were used. A subblock of 4 strips, 28 photographs, was se- 
lected for the test. Flight direction is eastwest,with 60 % forward 
overlap, 20 % side overlap. Each photograph contains 9 pairs of image 
tie-points, located approximately in the standard positions. 11 plani- 
metric control points are distributed along the perimeter of the block, 
at average distances of 2.3 base lengths (acutally irregular spacing 
between 1 and 3b). 17 vertical control points were used, of which 13 
are located at the perimeter (10 of them identical with planimetric 
control points) and 4 inside the block. For the accuracy investigation 
23 planimetric and 10 vertical check points were available, irregularly 
distributed over the block. The geodetic precision of all control and 
check points is known to be 1.5 cm in x, y and z. 
3.2 The block was measured twice in the analytical plotter Planicomp C 
100. In both sets the photographs remained unaltered in the instrument, 
respectively, and the same interior orientation was used, as well as 
the same reduction for lens distortion, refraction and earth curvature. 
-23 -