coordinates are corrected for system errors, rotated and transformed. 
Additional tie points may be chosen and measured with no relation to the data 
banks. As a result of these operations a new file of lateral tie points is 
gradually built up for the use in the final block adjustment. 
sr dete 
uy 
rl 
Data Banking System 
A suitable organization of collected data is not only important for the final 
off-line processing and block adjustment, but it is also crucial for the 
support of certain auxiliary functions which eventually make on-line 
triangulation so efficient. Several important data files are continuously 
the generated, stored and updated on disk during all operations. They are later 
ne used as data banks to search for and retrieve the already available 
ent information. In the progress of triangulation data banks are expanded by 
n appending new information to existing files. The data banks are accumulated 
and preserved for as long as needed for the triangulation block in progress 
and are cleared only by a request combined with a positive reconfirmation. 
An appropriate reorganization of data banks is implemented by a special 
program, after new measurements are entered in a restoration of an old model 
or part of the strip. The files are specialized to store independently photo 
and model coordinates of all points measured in the process of bridging, 
The measured coordinates of fiducial marks, transformation parameters derived in 
inner orientation, orientation elements of all photographs involved and photo 
coordinates of lateral tie points. 
rn roO D 
ier prs 
Quality Control 
An efficient quality control of observed data in every MODEL solution is 
implemented by applying the data snooping technique for gross error detection 
(Kratky, El-Hakim, 1983). Values of redundancy numbers and standardized 
residuals are computed for each observation, a rigorous statistical test is 
Dy applied and the results displayed. Weak points indicated by the test are 
remeasured and the test is repeated if necessary in a "trial and error" 
approach, which is extremely well suited for an on-line environment. The 
testing is accessed from the measurement section of the MODEL routine by 
manually activating switch A4 twice in a sequence (see flowchart in 
Appendix A-3). 
In this regard two important aspects must be considered - the accuracy of 
te collected data and the efficiency of used procedures. The efficiency is not 
s only judged by the time needed to perform the operations, but also by the 
intrinsic reliability of data. The on-line calculations are always applied 
to data of limited volume and are not so much important for the final, 
1x usually independent, block adjustments, as they are for the crucial function 
of quality control. One should take great care that statistical testing is 
not too adversely affected by the limitations of the on-line geometric 
configuration. This will ultimately be improved in the final, simultaneous 
processing of data, but then statistical tests are much more difficult to 
run. It is, therefore, important to consider suitable point configurations 
which guarantee high reliability of statistical testing at the time of the 
data collection and editing, rather than to try to reach the highest possible 
m accuracy of intermediate numerical solutions. 
ry 
> 
© © 
4 
Qu 
Ww 
(A 
Numerous practical experiments and on-line simulations showed that the 
critical areas for the testing reliability are in model corners. It appears 
to be very efficient to strengthen the stability by doubling points in these