105 ground control and/or check points were measured level of number of number of "ns
in the imagery and in maps of scale 1:25000. Their dis- pyramid interest operator conjugate iis
tribution with respect to the nadir imagery is shown in points points s
Figure 2 (the case with 20 ground control points is 4 796 552 115
selected). 3 2122 1700 REIE
2 5273 4265 “iis
1 11871 9109 115
DISTRIBUTION OF GROUND CONTROL CO) AND CHECK POINTS (+) DORFEN/20 GCPS 0 28359 14802 ns
115
3000 + Table 2. Numbers of interest operator and conjugate 115
e Ut g o is DA points on the five levels of the image pyramid
mods , ° "um ; ; o : for strip Dorfen (reduction in resolution by a
+ + toc factor 2°” in line and column directions)
2000 4 * * +
2 o t s : s. : 4.3 Error analysis Hs.
d 1568 + oy : e t 2 1433
S um. + tw s : no. Theoretical error measures: fas
iiu. m est , t mt s Jee, 1 To analyse the precision of the results of the adjustment E 1431
S02 t. + * a, Eo d v + 9 the theoretical standard deviations for the coordinates E unc
2 gt 0 : ; * o ; oat of all points given by the corresponding coefficients of v
2 toge 2600 — 3000 oe Le, Sue. TWO cum the inverse of the normal matrix were averaged sepa- F
rately for zones with twofold (zones 1 and 3) and ye
Figure 2. Distribution of ground control and/or check threefold (zone 2) stereoscopic coverage. These values 1427
points are given in Table 3 for a varying number of ground
number of 2
4.2 Details on photogrammetric adjustment zones Gc, (m) a,(m) 9, (m) ground control . P
points xi
The initial values of the exterior orientation were pro- 1 24 16 55 99 : ur
vided by the inertial reference system of the aircraft. 2 17 13 32 a 54
The exterior orientation of 120 orientation images have 3 24 16 52 E
been introduced as unknowns into the adjustment. Fig- ; : = ; + -54
ure 3 and Figure 4 show initial and final values of the all 2.1 1.6 4.4 3-12;
exterior orientation. 7 DELE
A crucial task in least squares adjustment is the 1 3.1 1.8 6.0 40 ad
selection of the proper weights for the error equations. 2 2.2 1.5 3.8 =
For this data strip the following weigths in terms of 3 2.9 1.8 5.7 D
standard deviations have been used finally: all 27 17 5.0
* collinearity equations: 4 pixels (42.8 im; included Fig
are - besides the image correlation errors - insuffi- 1 3.5 1.9 6.4 30
cencies in geometric calibration of the camera and 2 2.5 1.6 4.3
the errors introduced by the linear interpolation for 3 3.3 1.9 6.0
the exterior orientation parameters between the all 3.0 18 54
orientation images)
* coordinates of ground control points: 20 m 4 4.0 21 7.0 20
* interior orientation: 10.7 um 2 2.9 1.8 4.9
3 3.8 2.1 6.5
. exterior orientation: all 35 19 59
— position (x,y,z): 100 m, 100 m, 25 m sac
— angles (roll, pitch, yaw): 0.02 deg for all 1 4.9 2.4 7.8 10 | 55%
2 3.5 2 1 5.7 ^
* second order Gaufi-Markov process was not used 3 4.6 24 74 v sec
for Dorfen strip e TN
all 4.2 22 6.8 456
In a first adjustment run using only 1000 conjugate RET
points and relaxed weighting intial estimates for the 1 6.4 30 99 4
biases of the exterior orientation have been derived. 2 46 26 76
Later on, 12751 conjugate points have been introduced 3 59 30 94
into the adjustment and the runs were repeated with ; ; : 3
different numbers of ground control points, varying all 5.5 2.8 8.8
from 4 to 99. All these adjustment runs converged in
two iterations and took about 2 hours CPU-time on Table 3.
Average of the theoretical standard deviations for
IBM 3090, each.
the computed ground coordinates of the conju-
gate points given for various numbers of ground
A DEM was generated from the final ground coordi- control points: zone 1 and 3 are the zones with
nates of the conjugate points by interpolation in the twofold coverage, zone 2 is the area of threefold
irregular set of points resulting from photogrammetric stereoscopic coverage; the numbers of conjugate
adjustment. It is given in Figure 5. points were 4030 for zone 1, 5524 for zone 2, and
3017 for zone 3 (12571 points for all zones)
72
