), 3- and 6-ray points
, 6- and 9-ray points
d. Moreover, the rms
standard deviations of
parameters w, ÿ, À at
d. All accuracy figures
nal equation matrix.
[
S HRS)
e
o
0
RMS VALUE
Q
erent block configura-
nd orbit/attitude data
— N
o o 2
RMS VALUES H(2) [m]
A
ito
o
it block configurations,
it/attitude data
wre shown graphically,
idard deviations of the
rameters according to
) adjustments without
r-free position and at-
na 1996
titude data (A), the accuracy of point determination only
depends on the standard deviations of the image coordi-
nates, the number of conjugate points and the geometric
constellation of the ray intersections. The rms values are
3m in planimetry and 10 m in height.
Figure 6: Rms values pug for different block configurations,
ground control information and orbit/attitude data
Figure 7: Rms values py for different block configurations,
ground control information and orbit/attitude data
The planimetric and height accuracies decrease, if the po-
sition and attitude data are introduced with realistic stan-
dard deviations (case B). The rms values amount to ps
— 277 m and p; = 50m. They are too high to be shown
true to scale in Figures 4 and 5. The reason for these
unfavourable values is the poor absolute accuracy of 200"
(bias) and 0.7"/s (drift) of the observed attitude param-
eters w. For the single strip this accuracy cannot be im-
proved by the adjustment. Consequently, the rms value
pe amounts to 201”. Figure 6 is not able to show this
high value true to scale.
Figure 8: Rms values py for different block configurations,
ground control information and orbit/attitude data
Case C leads to a singular configuration because for bias
and drift of w no observations are available, and any other
observations are not able to determine the unknowns c.
Next the results of the block adjustments without GCP
are analyzed. The accuracies which are achieved for the
single strip (case B) can be improved considerably, if a
block with q — 2096 is adjusted (u$ 4 — 11m, uz = 20m).
The roll angles & are now determined with an accuracy
of pa = 6" due to the absolute position data, which were
introduced for each strip of the block. The accuracies are
only slightly poorer for case C.
In case of the block with q= 60% the accuracies of both,
object coordinates and orientation parameters are im-
proved by a factor of about 1.4 compared with the g = 20%
block. For the object coordinates u$; = 8m and uz
= 15m (case B) can be obtained, that is by factors 3.2
(planimetry) and 2 (height) poorer than in case A.
In the following the adjustment results using GCP are dis-
cussed. Figures 4-8 show that high accuracies are obtained
if 16 GCP are incorporated into the bundle adjustment. In
Figures 4 and 5 the rms values of cases B and C are close to
the ones of case A with error-free orbit and attitude data,
even in case of single strips (us $ = 4m and pg; = 11m
for B). Using the equal number of GCP the accuracies im-
prove slightly for the q — 2096 blocks and considerably for
the q = 60% blocks due to the increasing block strength.
The blocks with q = 60% provide rms values of u y y — 3m,
pz = 8m and pe = 1”, pe = 1”, px = 2" respectively, for
Band C.
Finally, it can be stated that all exterior orientation pa-
rameters are estimated with high accuracy, if a few precise
GCP are combined with position data of high relative ac-
163
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
