bul 2004 
location 
iL HRS 
| Terrain 
, called 
quisition 
location 
003] and 
emained 
ectations 
rence3D 
possible 
already 
d means 
ssioning 
ance for 
CNES), 
ıtegrator 
las been 
alysis of 
location 
tution's 
libration 
1 image, 
ntly and 
for each 
1e of tie 
ereopair 
al HRS 
process 
) ground 
different 
location 
sed to 
vements 
ification 
orithms. 
/e been 
ctivities 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
  
2.2 Available means of measure 
For the 1% interpretation level, analyses only concerned stellar 
location unit attitude restitution. In fact, different sources for 
attitude restitution are available and can be compared : the 
theoretical requested attitude, the stellar location unit attitude, 
the on board attitude control system (AOCS) which do not 
involve the star tracker, and finally the star tracker does its own 
attitude restitution. All these data are stored each time the 
satellite is in visibility of Toulouse, the main ground station, 
and available in a data base for further analysis. 
For the 2" interpretation level, system location monitoring data 
is used. It consists in a database of images on specific location 
sites all over the world, constantly updated : approximately one 
image for each SPOTS instrument is acquired each month on 
each site for this monitoring. Sites have been chosen for their 
cartographic data availability and also for their repartition all 
over the world in particular with latitude (see figure 3). 
"t e 
«cor . À 
A A 
Figure 3: SPOTS location sites around the world : circle = main 
site, diamond = secondary site, triangle = no more used 
Ground control points are taken on every image of this database 
and specific tools allow computation of the system location 
performance and also interpretation of the errors in terms of 
chosen parameters. Available parameters are polynomials up to 
degree 2 for attitude angles, bias for satellite position, bias for 
detectors viewing angles, focal length adjustment and potential 
rotation of CCD array in the focal plane. Thanks to the world- 
wide repartition of the sites, and the constant update of the 
database, analysis of errors dependencies with satellite orbit 
position and with acquisition time can be done. 
The 3" interpretation level has been performed in co-operation 
with French National Geographic Institute Space Department 
(IGN Espace) involved in the working group. IGN has put on 
some important means in order to ensure that Reference3D 
horizontal location performance is met. Analyses of bundle 
block adjustment process and errors for both monitoring sites 
and Reference3D production blocks has been performed. The 
large size of some production sites allows detection of specific 
behaviours which can not be seen in any other ways. 
Finally, when necessary, specific information about thermal 
behaviour, available thanks to on board thermistances, have 
also been analysed. 
3. ANALYSIS RESULTS AND IMPROVMENTS 
CARRIED OUT 
In this section, we will describe for each level of interpretation 
the main phenomena detected and studied and the 
improvements brought out of it. 
3.1 Attitude restitution analysis 
Stellar location unit attitude restitution is a new one on board a 
SPOT satellite and it had to be finely checked out because it 
determines the major part of HRS' location performance. 
Its high frequency performances have been assessed by 
comparison of image matching grids with location grids using 
the stellar location unit attitude data. It showed out remainders 
lower than 1 m (0.1 HRS pixel) when a disturbance such as a 
mirror movement is present, and usual microvibration lower 
than 50 cm (0.05 HRS pixel). See [Valorge, 2003] for more 
details about it. 
Comparisons between different attitude sources showed out an 
accordance between all of them, and highlighted several 
phenomena which do not put in question ULS' performances 
level but whose impact on the location performance has been 
studied. From it, two possible improvements have been 
identified. 
First, some roll perturbation of ULS attitude was present 
because of a deformation of the structure supporting the star 
tracker. This phenomenon has been removed thanks to an 
adaptation of the on board thermal control parameters. It 
allowed to gain a few meters on absolute location performance 
and time registration performance. 
Second, a possible improvement of the filtering process of the 
stellar location unit algorithm has been identified in order to 
remove measures gap due to confusion between stars and 
proton's impacts. This phenomenon only impacts a very small 
proportion of images and should soon disappear thanks to an on 
board modification. 
3.2 System location performances analysis 
Analysis carried on location performances measured for each 
HRS camera concerned both relative location (difference of 
location between the two HRS), and absolute location (mean 
location of the two HRS). 
At first sight, relative location appeared to be very noisy across 
the track. This was a problem because as measuring HRS' 
relative roll gives an estimation of the satellite yaw, it meant 
that satellite's yaw estimation was noisy too. This noise has 
been related to a deformation of the foot supporting the cameras 
and greatly decreased thanks to an adaptation of thermal control 
thresholds. 
Before the modification, the standard deviation for HRS relative 
location across the track was about 20 m, and it went down to 
7 m after (see figure 4). In the same time, the standard deviation 
for satellite yaw estimation out of HRS' relative roll went from 
100 microradians before down to 12 microradians after.