ympared.
entation
kewness
ed using
gnificant
what was
was not
, and the
onjugate
ction]
53D
1: 14.4
"| 14.5
| 14.0
1: 17.9
21.3
i2
on with
data],
tion]
3D
173
17 2
20.4
2277
22.3
Ón with
data],
Qn
e 5. The
he town
id Neto
id their
20 maps
S were
f which
defined
ible was
s, which
entation
Altitude 568km
No of CCDs per line 4 096
CCD size Zum
Pixel size 18.3m x 24.2m
Principal distance 213.5mm
Along track angle Imagel O°
Image 2? 15.
B:H 0.3
Table 5 - Characteristics of OPS sensor and data.
Results of tests with OPS data
The initial computation of the data set resulted in
problems of convergence, mainly affecting the
computation of the platform’s altitude for the physical
model. An exhaustive study resulted in the use of only
eight of the orientation parameters. It was necessary to
do this because of the small B/H which results in high
correlations between the orientation parameters. Also,
the given value for the eccentricity of the orbit was not
well defined by the literature. The only information
offered was that it should be smaller than 0.0015 in all
cases.
The errors presented were found in a very small number
of check points. Of the 40 ground control points, those
which were not used as control, were adopted for
checking the model’s accuracy. The accuracies found for
some of the tests are presented in tables 6 and 7, using
the physical and the polynomial orientation models,
respectively.
No |No |No rms (m) [in UTM projection]
iter |contr] conj.] E N H 2D] 3D
6 0 62 84 96 | 106 | 131
58 86 79 |1..104-]- 130
56 91 78 |-107 | 132
62 89 89 | 108 | 140
27 67 92 02 | 114 [| 150
SD DA DE
HR ja Un JON
ISIS
—
Table 6 - OPS model accuracy after orientation with
several control configurations, using the physical
orientation algorithm.
No |No |No rms (m) [in UTM projection]
iter [contri coni.] E N H 2D | 3D
6 0 71 86 97 | 113] 143
4 74 89 95 | 114 | 148
74 92 93} 118 } 150
8 72 9] 96 | 116 | 151
12 1785 103 | 101 | 134 | 167
oo |oo|Ox|-3|oO
CA [CA ION
e
Table 7 - OPS model accuracy after orientation with
several control configurations, using the polynomial
orientation algorithm.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
Similarly to the tests with SPOT data, the tests carried
out with the OPS along track imagery resulted in time
expensive solutions when conjugate points were adopted.
The physical approach for the orientation gave the best
results, both for relative and absolute orientations of the
model. However, the algorithms took more iterations to
converge for a minimum number of control points.
The results concur with other reports presented on the
study of the accuracy of OPS data in height [Maruyama,
1993]. The large errors found are most likely due to the
large errors in the control data and the problems found in
the identification of control. The small angle of
convergence can give rise to instability in the orientation
of stereo pairs. The B/H=0.3 is the major constraint for
the orientation of the data with this algorithm because it
is extremely influential on the ellipses of error, hence on
the correlations between some of the orientation
parameters.
Besides the savings in time spent in the orientation
process, it gives the same kind of accuracy with less
control points.
5. CONCLUSIONS
The use of conjugate points were tested with SPOT and
OPS imagery for the two different models. The
algorithm becomes very time expensive with the use of
conjugate points.
First, it introduces a few more equations to the
calculation, resulting in the inversion of larger matrices.
However, this would not be important if an improvement
in the final accuracy of the models was registered.
Second, the set of observation equations is not as stable
as when only ground control is used. A deeper study
showed that this is due to the different scales of the
residuals in the observation equations derived from
ground control and conjugate points.
No significant improvement in the orientation of the
model stems from adding conjugate points to the control.
It was also found that the convergence of the algorithm
depends on the number of observation equations formed.
The absolute orientation of the models gets worse when
the number of ground control is decreased independently
from the number of conjugate points being used.
However, conjugate points may be used to ensure a
model orientation when the ground control is not enough
on its own.
If a good initial approximation is obtained for the
orientation parameters, conjugate points can be used with
success to improve the relative orientation of the three-
dimensional models. However, the results obtained with
this study were not significant.
