- T
The condition numbers in table 1 confirm the results from
chapter 4, On the other hand, although the correlation values
for testfield 1 are already very high (see /1/, table 4), the
condition numbers are nevertheless acceptable.
From now on we compute condition numbers and correlation values
with the complete set of informations ; i. e. the equations of
type 3 in chapter 3 for the orientation angles will be added
to the design matrix A of table 2 in /1/. The results are shown
in tabel 2.1 - 2.5. From obvious reasons it is of essential
importance for a calibration procedure to produce the set of
calibration data with as little correlation to the rest of
parameters (not realy wanted) as possible. Therefore, in the
tables only these critical values are shown. Whether the other
correltion values are high or not can be estimated from the
condition number. Examining tables 2.1 - 2.5 the following
characteristics may be evaluated:
- Introducing the readings of orientation angles strengthens
the mathematical model coneiderably: the condition numbers
are improved and the generell correlation level is lowered
(see tabel 1, 2.1 - 2.3 and in /1/ table 4).
- Condition numbers and correlation values do not differ very
much in all three testfields. Both sets of values improve
from testfield 1 to 3 but not considerably. So, even a test-
field consisting of only one row of points is sufficient for
a complite camera calibration.
6. CONCLUSION
As a final consequence the here proposed and with some tests
investigated camera calibration method turns out to be numeri-
cally stable and capable of a reliable evaluation of the wanted
calibration data. Since in the extreme case only one row of
terrestrial points have to be targetted and geodetically
measured this method is economic and efficient as well.