3. The practical performance of the distortion determination
In practice we cannot assume the photogrammetric model to be
exactly oriented in any stereoscopic instrument. We therefore must
compute the final relative orientation from the residual parallaxes,
which remain in the orientation points after a preliminary orientation.
I'rom the obtained corrections to the elements of the relative orien-
tation, corrections can be computed to measured y-parallaxes in all
points of the model. See for instance [2]. After the corrections have
been applied to the measured y-parallaxes, the residual parallaxes can
be used for the distortion determination as described under 2. For the
computation of the corrections to the elements of the relative orienta-
tion, the measured y-parallaxes in the points 1—6 can be used. In addi-
tion the points /—9 can also be used.
It is also possible to express the residual y-parallaxes p, after the
r
adjustment as direct function of the measured y-parallaxes before the
adjustment of the relative orientation. See eq. (9) below.
The standard error of the y-parallax measurements will become more
reliably determined, if the mentioned nine points are used for the ad-
justment of the relative orientation instead of only six points.
In summary the procedure is:
l. In the contact prints of the negatives in question or on the draw-
ing table, the points for parallax measurements are indicated according
to fig. 1.
2. After a preliminary relative orientation in the instrument the y-
parallaxes are measured in the indicated points. Stereoscopic measure-
ments should be used in order to obtain high accuracy of the measure-
ments. The measurements in each point should be performed in at least
three details.
The standard error of the y-parallax measurements in the projection
plane is computed from the formula
1 / D 7 = : ^ (8)
Ac: V a 2p; + 2p; c ps — Ps U Ps — po)? t
T6] (ps + Pi Z2p7)? + (Ps t pe — 2ps)? + (pi t pe — Pa)
3. From the measured y-parallaxes in the points 16 and all points,
the residual y-parallaxes p , are computed according to the formula:
ST
I
Be (4p + 2P2 + P3 — ps t ps — pg) +
D
x
m 3b (—Pı P03 5 ——p +p) +