^ic 
13, 
a- 
ned 
is 
1,4 in height, Because of the limited accuracy of the contro] 
points and check points, by which the accuracy figures y are 
distorted the more, the smaller they are, the real accuracy 
advantage of bundle block adjustment can even be expected as 
higher, 
- Table 4 - 
Altogether the test results demonstrate, that by simultaneous 
self calibration excellent accuracies can be obtained, even if 
systematic errors of considerable size are existing. As an 
example we cite the RMS values ux,y obtained with the extreme 
control distributions i. = 8 and.i.-= 11. Here, both adjustment 
methods lead to amounts of 7 um to 8 um at the photo scale or 
20 cm to 22 cm in the terrain. If we compare these accuracies 
with the control spacings of 20 km to 31 km we obtain ratios 
which are better than 1:105, 
Finally, the important statement can be made, that the test 
results, obtained with simultaneous self calibration meet the 
theoretical expectations in a twofold way. Firstly the standard 
deviations of unit weight co, oop and coop are practically inde- 
pendent of the control distribution and secondly the empirical 
ratios wog, representing the error propagation with the block 
adjustment, are in well agreement with the corresponding theo- 
retical predictions, being based on random errors only. These 
facts indicate,that the systematic deformations of the image and 
model coordinates are extensively compensated and that the re- 
maining errors can be considered as random. 
5. CONCLUDING REMARKS 
This reduction of the data errors to the purely random component 
is the most important result of the test. It confirms, that the 
used strategy for self calibrating block adjustment is fully 
effective in the present case, For a real] conclusive valuation 
of the suggested Strategy however, further and more generally 
drafted tests with block variable systematic errors and diffe- 
rent overlap configurations still have to be performed. 
APPENDIX 
Self calibrating block adjustment with additional parameters, 
treated as a collocation problem. 
Let us first formulate the functional model of block adjustment 
as: 
Ax - f = 0 2 (2) 
f is the observation vector, containing the measured image or 
model coordinates and x is the vector of the unknown terrain 
coordinates and transformation parameters. By X an f the theo- 
retical values of x and f are ment. 
The mathematical model of collocation supposes, that the actual 
observation vector f differs from the theoretical one f due to 
two random vectors n and s with the statistical expectations 
zero (see Moritz |16]).