E. Accuracy checks 
While the work described in Section D was being performed by the Survey Bureau 
of Dr. Schlund, tests of the results obtained were carried out by the Survey Section 
of the Swiss Federal Railways, under the leadership of Dipl. Ing. ETH P. Fülscher, 
partly in collaboration with the contractor. 
1. A first check on the accuracy of the photogrammetric process was given by the 
residual errors in the passpoints after Helmert transformation but before adjustment. 
The result was + 3 cm, but this value contains the error of the ground survey, which 
is of the same order as the error of the photogrammetric survey. Therefore the mean 
3 
V2 
square error of the geodetic and photogrammetric coordinates may be taken as 
+ 2,1 cm. All 6 models gave the same accuracy. 
2. A total of 80 points was measured in neighbouring models. The mean square value 
of the differences between the two measurements was + 3 cm (maximum error 9,8 cm), 
corresponding to a mean square point error of > = + 2,1 cm. The result was the same 
for all overlapping models. If the differences of coordinates of the twice-measured 
points are plotted graphically it is seen that in most cases there is a systematic error, 
caused partly by the accidental errors of observation in the passpoints, partly by 
local film shrinkage and partly by lens distortion. The average value of these syste- 
matic and accidental errors was approximately 7 um in the negative. No correction 
plates were used in the Autograph. 
3. The redundant curve-points could also be used as further check points for the ac- 
curacy of point coordinates with the help of a computer program of the Swiss Fede- 
ral Railways for computation of radii from coordinates. The distances of these points 
from the unadjusted centres of curvature were computed. The mean square distance 
error of approximately 3 cm computed from these results is only of limited value, 
since it contains all the large errors of track displacement and the inaccurate determi- 
nation of the tangent points. In fact an essentially higher accuracy must have been 
achieved. 
For practical work the distance error between neighbouring points is more important 
than the absolute coordinate error. The mean square distance error was computed 
from the comparison between the lengths derived from the coordinates and the check 
lengths measured in the field or from the known standard lengths in railway switches. 
4. 41 control lengths were measured with a tape in the second model (distribution 
zone). Since only a few of these lengths stretched from one model to the next there 
was no opportunity for a complete check of error of transfer between models; the