A minimum of two control points per image is advisable
for a good absolute orientation of the model. However,
more control is needed for an orientation considering all
the parameters. Two ground control points give eight
observation equations for each image of an across track
stereopair. If only conjugate points are to be added to this
set of control, a minimum of four points is required to
obtain the four lacking equations. For the case of along
track imagery, two ground control points give four
obsevation equations, and an extra eight conjugate points
are necessary to obtain the eight more equations needed.
4. RESULTS OF TESTS
4.1 Tests with SPOT data
SPOT data
The SPOT data used covered an area in South East
France, used in earlier orientation studies [Dowman
et.al., 1991]. The main characteristics of the data are
summarised in table 1. A set of 106 ground control is
known for the area, and on-board registered data is
available.
Altitude 830km
No of CCDs per line 6 000
CCD size 131m
Pixel size 10m x 10m
Principal distance 1082mm
Across track angle Image l 22.38
Tage 2. 20.5
B:H 0.8
Table 1 - Characteristics of SPOT sensor and data.
Results of tests with SPOT data
Several tests were carried out with the SPOT data for
different control configurations. Table 3 summarises
some of the results obtained using the physical
orientation model, with an indication of the number of
control and of conjugate points used. Some of the results
obtained using the polynomial model are summarised in
table 4.
The use of the on-board registered data gives a good
initial orientation model, reducing the number of
iterations necessary. Although worse results were
obtained using the polynomial approach, it still proved to
adapt well to the short arcs of the orbit. However, the
polynomial algorithm did not converge where less than
three control points were used.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
The final SPOT models were analised and compared.
Although a slight improvement in the relative orientation
characterised by smaller errors in the skewness
parameters was observed for the models oriented using
conjugate poi nts, this improvement was not significant
when compared to the errors found. Contrary to what was
initially expected, the relative orientation was not
significantly improved in the case of SPOT data, and the
algorithm becomes more time expensive when conjugate
points are used.
No |No |No rms (m) [in UTM projection]
iter | contrfconj.| E N H 2D | 3D
4 6 0 197 | 83 10.6 | 128} 144
2. 6 0° 19% | 183 | 60.5 P129] 14,5
S" 6 4*1 96 | 79 64 | 124] 14.0
6* 4 85142.0 1,010.53 |.83.1 | 15.9 |. 17.9
7* 2 12 T1341 141 |] 8.7.1 194; 21.3
Fi 2 16.1132.1143 1.8.8 1.10.5.1,21
Table 3 - SPOT model accuracy after orientation with
several control configurations [*using header data],
using the physical orientation algorithm.
No |No | No rms (m) [in UTM projection]
iter |contr| conj.] E N H 21D 1 3D
4 7 0 "10:9 ETS 107 1° 1158} 17,3
4* 7 2 (108) 14,570 | 158-172
4% 6 5 [123] 14.1 | 8:3 1 18.7 120.4
S* 4 10 1144 | 44.5 | 99 | 2041 20.7
6* 4 16 |14.4 | 14.1 | 9.6 | 20.2 | 223
Table 4 - SPOT model accuracy after orientation with
several control configurations [*using header data],
using the polynomial orientation algorithm.
4.2 Tests with OPS data
OPS data
A summary of the OPS data used is given in table 5. The
data covered an area in the French Alps around the town
of GAP. As earlier reported by Dowman and Neto
[1994], 40 control points on the two images and their
ground coordinates were extracted from 1:25,000 maps
of the region. However, many problems were
experienced during the identification process of which
most were related to the difficulty of finding well defined
points on the imagery. The ground control available was
concentrated over the area in three main clusters, which
is not an ideal control configuration for the orientation
process.
Z|
©
