nbul 2004
)3 03-04
03 03-04
03 03-04
trend
1ance
analysis
pairs, and
have been
3eographic
2003] and
| account a
of the on
all images
These are
ossible for
dle block
iween two
{RS strips.
ering HRS
anage it. It
lo not vary
d allows a
RS inner
lar, among
s available
| length.
cation site
sites (total
ed out that
key for a
riority was
d of some
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
4. HRS' CURRENT LOCATION PERFORMANCES
In this last part, we will first give the latest figures about HRS'
location performance available from the ancillary data provided
with the images. Second, we will give a breakdown of errors
sources making up the performance.
4.1 Latest figures for HRS' location performances
Figures given in this paragraph concerns HRS latest
performance assessment for year 2004. Two types of figures are
presented : Table 1 shows the relative location performances
(difference between the two HRS' cameras location) and
Table2 gives the absolute location performance (mean
performance for the two cameras).
Statistical are computed out of location difference and mean for
each HRS stereopair. For a single stereopair, location difference
and mean are computed out of mean location of each image of
the stereopair.
HRS relative location
performances
(77 pairs)
meters | | track | // track | Global
Min -16 -39
Max 19 29
Mean 4 -5 6
Standard deviation 8 10 13
Max for 90% of pairs| 14 | 14 20
Table 1 : HRS relative location performances
(with no tie points used)
HRS absolute location
performances
(77 pairs)
meters | L track | // track | Global
Min -20 -42
Max 19 19
Mean -2 -2 3
Standard deviation 9 12 15
Max for 90% of pairs | 15 | 18 27
Table 2 : HRS absolute location performances
(with no ground control points used)
4.2 Error sources affecting HRS location budget
In this paragraph, we'll give the size and nature for each
elementary source of error making up HRS' absolute location
performance. All error sources are summarised in Table 3.
Geometric calibration remainders are high frequency errors
varying in the field of view. Their contribution should be
smaller than 1 meter rms in both directions (across and along
track) thanks to the high accuracy of reference data used to
perform the inner orientation (see [Gachet, 2004]).
Doris ephemeris provided with the images have an accuracy
better than 1 meter rms in both directions (along and across the
track). Ephemeris errors are mainly bias depending on the
image and can be lowered by ensuring ephemeris join for
different images acquired from the same orbit.
Image line dating accuracy is a low frequency error : it is
mainly a bias depending on the image (accuracy of the center
line dating) combined with a very few linear dependence due to
line period measure precision. Its contribution is about 3 meters
rms and only impacts along the track.
Structure deformations of HRS, or between HRS and stellar
sensor unit are supposed to be low frequency errors as an
hypothesis for our analysis. The temporal trend observed (see
paragraph 3.3) allows to evaluate this error about 8 meter rms
across the track and 5 meters rms along the track.
Finally, attitude's restitution accuracy delivered by the stellar
location unit combines both low frequency and high frequency
contributions due to its specific nature (see paragraph 1.2 and
3.1). Given all other error sources and the final location
performance measure, its accuracy should be better than 5
meters rms across the track and 10 meters rms along the track
which is in accordance with expectations.
HRS absolute location
budget
meters | L track | //track | Global
Geometric Calibration 1 1 1
Ephemeris 1 1 1
Line Dating 0 3 3
Structure 8 5 10
Attitude restitution 5 10 10
Table 3 : HRS absolute location contributors
(root mean square figures)
These figures are deduced on the one hand from our knowledge
of each error source and on the other hand from observations
and analyses carried out. As potential structure deformations
are still not very well known and cannot be precisely measured,
the exact repartition between structure and attitude restitution is
linked with our current interpretation of available measures and
may be subject to some evolution if this interpretation happens
to change.
5. CONCLUSION
The joint working group activated after the end of the
commissioning phase in order to improve HRS' location
performance has achieved significant results. Several
improvements were brought, and objectives were met : HRS
location performance is now about 15 m rms and compliant
with Reference3D planimetric specification.
Now the main contributors left are closely monitored on a
monthly basis. This constant monitoring of location accuracy is
coupled with a close attention paid to the on board stellar
location unit in order to detect every potential problem as soon
as possible.
Further work is still planned with HRS in the context of its
monitoring. First the temporal trend identified needs to be
constantly evaluated as it is taken into account into HRS world-
wide database Reference3D's production process. Second, it is
planned to test and compare different inner-calibration of HRS
on different sites in order to improve the inner orientation
parameters and validate the hypothesis of time stability of these
parameters.