Following this, the computation of mean square values of the residuals has been
repeated for the remaining ground control points. These values are listed in Table YI6.
Tables AAa and VIb show that with second-degree adjustments, the best result
has been obtained with the Zeiss Jena measurements, while the Wild A7 results are
the poorest. The same conclusion had been drawn already from the mean square
values of the relative errors in the tie points, listed in Table IV.
However, in the case of corner control only, the Wild A7 results are the best.
As shown by the graphs with the residuals in the ground control points, this is caused
by the larger systematic errors in uncontrolled areas for the Zeiss Jena and the Nistri
TA3 adjustments.
Table VIb indicates that for the specified purpose of the test, which is mapping
at the scale 1:100000, each of the ten adjustments gives a satisfactory accuracy.
The adjustments on perimeter control give an acceptable accuracy for plotting at the
scale 1:50 000, and the adjustments on control along the western and eastern sides
are only slightly less accurate than those on perimeter control.
The two third-degree adjustments with corner control are somewhat less accurate
than the corresponding second-degree adjustments. Apparently, the greater flexibility
of the third-degree adjustment entails a greater possibility of introducing unwarranted
deformations and, thereby, systematic errors.
e) Best possible accuracy of the block adjustments
In addition to the above results, it is of interest to know what the best accuracy
is that could possibly be achieved by means of the method of second-degree conformal
transformation of strips. This accuracy can be computed in two ways: 1) by using
all available ground control points for the computation of transformation formulas
for individual strips, and 2) by not increasing the number of ground control points
in the block adjustment but ensuring that no perceptible identification errors occur
in the used points.
Accordingly, each strip has been transformed using all ground control points
that have no obvious error larger than 10 m. In addition, the block adjustments with
perimeter control have been repeated after replacing a number of control points that
have errors not representative of those in surrounding points. The following points
have been replaced:
Zeiss Jena
Nistri
41 by
42
5
by 4
3 (missing)
by 4
192
191
19
20
271
by
270
118
193
41
37
271
273
192
119
271
265
297
298
In each case the mean square values of the residuals have been computed. These
are listed in Table VII.
36