33
erlapping
cal shape
that the
? the pre-
y-parall-
tlie most
jsults are
3.2 Practical Tests of the Precision and Accuracy of the Relative Orienta
tion in Stereoscopic Projection Instruments
The precision of the relative orientation is expressed here in terms of the precision of elimination of
the y-parallaxes. This precision is determined from replicated or repeated corrections of the y-parallax
in individual points of the model and is expressed as the standard deviation of one correction of the
y-parallax. The standard deviation is computed from the deviations between the average of the readings
from replicated or repeated corrections and the corrections themselves.
The accuracy of the relative orientation refers to the discrepancies in the requirement that all pairs
tade. See
rnational
tunity to
Results of
ilts from
of rays shall intersect in space after the relative orientation. This means that the discrepancies are mea
sured in a number of representative model points. In order to increase the precision of the measured
discrepancies (y-parallaxes) the correction in each point is replicated or repeated a number of times
(usually at least three times) and the average of the readings is computed and used as the measured
value. Then the accuracy of the relative orientation can be expressed in terms of the root mean square
value of the measured y-parallaxes or, if an adjustment according to the method of least squares of the
measured y-parallaxes is made, as a standard error of unit weight of the y-parallaxes. This latter value
will define the highest possible geometrical quality of the relative orientation under the circumstances
present. It is known that the discrepancies of the relative orientation (the y-parallaxes in selected model
points) can be used for a determination of certain regular errors in particular radial distortion of the basic
image coordinates and projectors.
detailed
different
As to the precision of the y-parallax corrections, there are some investigations reported from the
organization OEEPE. See reference 3.2:1. The standard deviation of one correction was found to be
4—6 /¿m on the scale of the image.
y-parall-
easuring
As part of the international controlled experiments of Commission IV I.S.P. (the working group IV:1)
a number of tests of the precision and accuracy of y-parallax corrections were made and the results are
summarized here.
ft image
dence is
In these experiments after completion of the relative orientation the residual y-parallaxes were mea
sured in at least 15 points of the models by different operators. The corrections in each point were repeat
ed at least three times. The precision was then determined according to the usual formulas.
The following values were calculated.
»bserved
parallax
Standard deviations
One y-parallax
Instrument correction Average
an from
rwenty-
•rection.
itermin-
Stereoplanigraph C8 9.9 microns 5.7 microns
Stereometrograph 4.7 2.7
Autograph A8 4.3 2.5
Autograph A7 7.9 4.5
aarallax
i devia-
tandard
lei tests
Average 6.7 3.8
The corresponding values for the averages of three corrections were obtained after division by \ 3 =
1.73.
The accuracy expressed as root mean square values s m i 5 of the residual y-parallaxes in 15 points and
as standard errors of unit weights s„i5 after adjustment was as follows.
ion me-
skillful
5 better
»ortant.
Instrument s m i5 s 0 is
Stereoplanigraph 10.3 microns 8.6 microns
Stereometrograph 12.1 8.9
Autograph A8 14.8 6.1
Autograph A7 10.5 5.6
ion in a
erection
'ections
From previous tests (see reference 1.342:1) the root mean square value of residual y-parallaxes was
found to average 12 ^m. The residts shown above are in good agreement with this. In the previous tests
mentioned, the standard error of unit weight was found to average 6 «m. There is good agreement with