Ph. KUBIK, A. MEYGRET, E. BRETON, F. MASSON
M. PAUSADER, D. LEGER, L. POUTIER
O. HAGOLLE, A. MEYGRET
M. DINGUIRARD, D. LEGER
F. VIALLEFONT, R. GACHET
AC. DE GAUJAC
P. HENRY, X. BRIOTTET
M. DINGUIRARD, D. LEGER
AC. DE GAUJAC, P. GIGORD
G. BEGNI, B. BOISSIN
M. LEROY, D. PRADINES
M. DINGUIRARD, D. LEGER
V. RODRIGUEZ, P. GIGORD
JP. DARTEYRE
+ C. LATRY, V. PASCAL
F. CABOT, F. DE LUSSY
40 years of experience
with SPOT in-flight Calibration
C. VALORGE, A. MEYGRET, L. LEBEGUE, P. HENRY (CNES)
A. BOUILLON, E. BRETON, R. GACHET (IGN) Page 1/30
International Workshop on Radiometric
and Geometric
2-5 Dec 2003, Gulfport,
MS
D. LEGER,
F.Calibration,
VIALLEFONT
(ONERA)
Overview
 SPOT system overview


SPOT satellites
SPOT system
 Geometric calibration and quality assessment



Geolocation model and accuracy
Internal orientation
Image deformation quality assessment
 Radiometric calibration and quality assessment



Radiometric model
Normalization
Absolute calibration
 Spatial Resolution


Refocusing
MTF assessment
 Summary
Page 2/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
SPOT system overview
 First generation :
2 identical instruments called HRV: 10m Pan - 20 m multispectral; steering mirror (+/-27°)

SPOT 1: launched 22 February, 1986
put on a 560 km orbit in November 2003. Re-entry in 2019

SPOT 2: launched 22 January, 1990
no more on-board recording since October, 1993

SPOT 3: launched 26 September, 1993
failed on 14 November, 1996
 SPOT 4: launched 24 March, 1998



New platform, same resolution
New 20m SWIR band (HRVIR)
VEGETATION payload
 SPOT 5: launched 4 May, 2002


Resolution (HRG):
5 m in panchromatic mode, 10 m in spectral mode
2,5 m in panchromatic mode through processing (THR)
Passengers: VEGETATION-2, HRS (High resolution stereo camera), Stellar Sensor
 Current operational constellation: SPOT2, SPOT4 and SPOT5
 Cumulated life on-orbit: 16 + 14 + 3 + 5.5 + 1.5 = 40 years ;-)
Page 3/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
SPOT system overview
 System description:


832km Sun-synchronous orbit; 26 days repeat cycle
900 km wide corridor, daily access
Operational architecture
Satellite
Operations and
Control Center
Network of Direct Receiving Stations
Image Quality
Expertise Center
Programming
Center
Processing and
Archiving Center
Page 4/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
SPOT system overview
 The Image Quality Expertise Center is in charge of
all in-flight activities regarding Image Quality:




Determination of the optimal on-board parameters (focus, radiometric
gains, compression parameters…)
Elaboration of the best ground-processing parameters (normalization
parameters, interior orientation…), validation and transmission to the
processing stations
Periodic assessment of Image Quality Budgets (wrt specifications):
« SPOT Image Quality Performances » issued every year, edited by
Spotimage, provided to their customers
Analysis and resolution of any image quality problem occuring in-flight
 This implies specific capacities:



Dedicated programmations of the payloads (even non-nominal)
Management of calibration sites and means
Dedicated facility, computers, operational interfaces…
 This Center is operated by technicians and engineers from
CNES, IGN and ONERA (up to 20 people during
commissioning phases, 5 for routine operations)
Page 5/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Geolocation model Calibration:

GCP database





Established since SPOT1, refurbished for SPOT5 (improved resolution,
need of better accuracy)
Planimetric precision: better than 5m for most GCPs
Sites covering at least 120 km x 120 km (HRS) + France
Special emphasis on the scattering of the location sites around the world
One bundle block adjustment per calibration site involving all
calibration acquisitions







Systematic programming of both HRV/IR/G + HRS
Between 10 and 20 GCP per image
More than 100 images per site in routine phases: robust estimations
Possibility to correct for erroneous GCP coordinates
Identification of correction parameters for each acquisition
Yaw, pitch and roll biases are then analysed in terms of calibration, for
each instrument, with respect to the steering mirror position, latitude, ….
Each new acquisition over a given location site is then added to the
corresponding block
Page 6/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Location sites over the world
Main sites: 12
Secundary sites: 4
No more used
Page 7/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Steering mirror viewing model:
mirror’s axis wedging
defaults
levelness default between mirror
axis and mirror plane


 
2   
Nr  Nn    Nn  2  m  sin  .Y
4
mirror ’s pointing
angle
Normal for a
perfect mirror
500
300
400
200
// track location (m.)
// track location (meters)
Real normal
300
200
100
0
-100
100
0
-100
-200
-300
-30
-20
-10
0
10
mirror angle (degrees)
20
30
-30
-20
-10
0
10
20
30
mirror angle (degrees)
Page 8/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Exterior orientation calibration:

After steering mirror calibration: remaining errors translated into biases
between instrument and AOCS reference frames

Interest of the world-wide scattering of our sites:

pitch (microrads)
150
100
50
0
-50
-100
-150
100
100
50
0
-50
-100
-40
0
40
-80
80
In yaw
-40
0
40
80
0
-50
-80
Satellite orbital position (degrees)
Satellite orbital position (degrees)
50
-100
-40
0
40
80
Satellite orbital position (degrees)
150
100
In pitch
In
100
50
50
0
0
International Workshop -50
on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
-50
ch (microrads)
-80
w (microrads)
yaw (microrads)

We quickly discovered an orbital variation of these « biases », the same
for each instrument on-board SPOT5
After analysis, due to a wrong reference date in the stellar sensor…
Constant biaises after correction of this on-board problem
roll (microrads)

roll
Orbital trends before correction
Page 9/30
 Geolocation accuracy assessment:
along track location (meters)
Geometric Calibration
80
40
Done simultaneously with calibration activities
0
Stringent SPOT5 specifications concerning geolocation accuracy:


50m CE RMS for HRS
-40
15m CE90 after bundle block adjustment without GCPs for Reference3D


=> intensive routine monitoring:
at least, each site must be acquired during -80
each -40
repeat0 cycle
days)
40 (2680
quasi-real time exploitation of these images across track location (meters)


80
along track location (meters)
along track location (meters)
-80
40
0
-40
-80
-80
-40
0
40
across track location (meters)
80
80
40
0
-40
-80
-80
-40
0
40
80
across track location (meters)
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Page 10/30
Geometric Calibration
 Internal orientation:

Absolute method


panchromatic and multispectral reference bands
(HMA and B2)
each HRS band

comparison image / reference

Relative method



THR mode (HMA/HMB bands)
Multispectral mode (B1/B2/B3/SWIR bands)
Relative panchromatic/multispectral (P/XS)

comparison of pairs of simultaneous images

Quality assessment:

Made simultaneously: same methods, different acquisitions, checked on
corrected images
Page 11/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Absolute internal orientation calibration:

Reference data: Manosque test-site




Aerial cover of a 60 km  7 km area at
1.50 m resolution with 80% overlap
Triangulation: 0.40 m. accuracy
Aerial digital surface model at 1 m
resolution
Method




each aerial image is projected into SPOT5
image geometry (taking the MTF into
account)
a fine image matching process measures
differences between SPOT and the
Reference
Filtering and averaging to get each
detector orientation
Final modelling of these curves:



Reference
Drift of along-track orientation = yaw
Drift of across-track orientation =
magnification
Higher degree tendancies = distortion
SPOT 5
Page 12/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Absolute internal
orientation calibration:

Corrections achieved:




Magnification of each
instrument
Relative yaw:
HRS1/HRS2,
HRG1/HRG2 and P/XS
Optical distortion: up to
fifth degree polynomials
After calibration:
residuals < 15 cm RMS
(limitation due to the
reference)
Page 13/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Image deformation quality assessment:

Dynamic perturbations monitoring
Dynamic perturbations: inertial wheels, magnetic tape recorder (SPOT14), steering mirror, attitude restitution errors
Specific programmations: phased pairs (26 days time lag => same
viewing conditions), Image Quality mode (simultaneous image of both
HRV/IR/G), autotest (SPOT5, mirror in auto-collimation position)
Dense image matching + line-wise averaging => profile vs time
First conducted on SPOT1 as technology experiments (87)
Nominal activity since SPOT2





residual roll shift (pixels)
roll shift (pixels)
1
0.15
0.1
0.5
0.05
0
0
-0.5
-0.05
-1
-0.1
-1.5
-0.15
0
20
residual roll shift (pixels)
0.15
40
Time (second)
60
80
0
20
40
Time (second)
60
80
Use of a phased pair to determine the influence of a steering mirror move
Page 14/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
Use of a IQ couple to determine the steering mirror stabilization time
Autotest pattern
Use of the autotest for the same purpose: only one acquisition during night
Page 15/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Geometric Calibration
 Image deformation quality assessment:

Length distortion assessment:



Conducted along with geolocation activities, using the same GCPs
Computed for each pair of GCP by comparing real/modelled distances
Analysis as function of orientation and length
distance error (meters)
50
40
30
20
10
0
0

Planimetric accuracy assessment:




10000
20000
30000
Distance (meters)
40000
50000
Location accuracy after bundle block adjustment with GCP
Can be assessed with high precision GCPs (residual analysis)
Can be assessed along with altimetric accuracy (need for reference
DEM)
Altimetric accuracy assessment:




Performed by value-added producers: IGN & ISTAR
Operational production capacity: optimal conditions, completeness…
Crucial point = reference DEM
HRS SAP initiative under ISPRS framework
Page 16/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Radiometric model:

Push-broom sensors => same model for all HRV, HRVIR, HRG, HRS
C(k,n,b,m)
A(k) g(k,n,b)
g(k,b)
G(m,k)
R()
8-bits
ADC
Rad
Amplification
Optics &
Read-out
filters
register (b)
Detector (n)
X(k,n,b,m)
X(k,n,b,m)=R[A(k).G(m,k)g(k,n,b).g(k,b).Rad(k,n,b)+C(k,n,b,m)]
Absolute calibration
Normalization
Page 17/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Normalization

Normalized digital count:
Y(k,n,b,m) 

X(k,n,b,m) C(k,n,b,m)
 A(k ).G (m, k ).Rad ( K , n, b)
g(k,n,b)g(k,b)
Dark currents calibration:





Steering mirror in auto-collimation position (HRV, HRVIR, HRG) or nightacquisitions (HRS)
Obtained for each detector of each spectral band with each amplification
gain by averaging its digital counts
Short term variation monitoring: 10 minute images
Medium term variation monitoring: one acquisition per week, then per
month
Difficult case: SWIR band (high increases due to proton collisions)
=> updated every week
Page 18/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Normalization

Inter-detector coefficients calibration
Problems
Antarctic with on-board lamp (steering mirror positioning accuracy)
Use of quasi-uniform landscapes: snowy expanses
Groenland


0°
Cote
70°
80 °
S4 ANTARCT_2002
S5 ANT_1202
70°
80°
C
7
8
S
H
EST
OUEST
180°
Antarctic (winter)


Greenland (summer)
Operationally heavy: 10% success (wheather, non-uniformity)
Correction of Solar incidence before averaging
Page 19/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Normalization quality assessment:


Made on uniform, normalized images: average line
Different criteria:





High Frequency
Low Frequency
Inter-Array
Even-Odd detectors
IA
LF
< 0.3 % for each
E/O
HF
Page 20/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Signal-to-Noise Ratio (SNR) assessment:
Column-wise Noise:



Use of on-board lamp (SPOT1-4)
Use of quasi-uniform sites with 2 simultaneous acquisitions (to separate
instrumental noise from landscape signal)
Noise model:





Physical understanding of noise sources (signal noise, digitization…)
Simple model:   a  K .Rad
Allows comparison of sensors in a common reference configuration
Line-wise Noise:
cf. normalization quality
assessment
Image-Noise:
combination of the two
previous noises
Column-wise noise for SPOT4 HRVIR2 M, may 2003
1,0
0,9
Column-wise noise (W/m2/sr/mm)

0,8
0,7
c(Lref,G2)
0,6
G1
G2
G3
G4
Modele G1
Modele G2
Modele G3
Modele G4
0,5
0,4
0,3
0,2
0,1
Lref(B2)
0,0
0
20
40
60
80
100
120
140
160
180
200
Radiance (W/m2/sr/mm)
Page 21/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration:


Operational use of many different methods
Important for users, but also for amplification gain prediction G(m,k)


Gain calibration using IQ mode
SPOT Histogram DataBase (started in 87)



Stores every cloud-free scene histogram on a 120km x 120 km grid
Gives statistically significant estimation of observed radiances
Monthly average used to predict optimal amplification gains
Page 22/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration: on-board calibration systems

Easy to use => frequent estimations
On-board lamp: not absolute but temporal variations monitoring
15
Sensitivity variation m easured w ith the
lam p
10
Variation (%)

5
0
-5
24/3/98
6/8/99
18/12/00
2/5/02
14/9/03
-10
-15
-20
-25
-30
-35
HRVIR1 B1
HRVIR1B2
HRVIR1B3
HRVIR1SWIR
HRVIR2 B1
HRVIR2 B2
HRVIR2 B3
HRVIR2 SWIR
Date
Page 23/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration: on-board calibration systems

Sun-sensor




Optical fibers (48 for HRV, 24 for HRVIR) projecting solar radiance onto
some detectors of each spectral band
Highly difficult to characterize before launch
On-orbit variation of fibers transmission
Successful only for SPOT4, abandoned for SPOT5…

u(t).TFIB(j) 0 E0().TBE().sk ().d
Lk 

o(j)
 sk ().d
0
S POT4 HRVIR1 B1
Numerical level
Eo(): spectral solar illumination
u(t): Earth-Sun distance variation
o(j): solid angle of fiber j
TFIB: transmission of fiber j
TBE(): spectral transmission of the
calibration unit
Sk(): spectral sensitivity of channel k
150
100
50
0
1
501
1001
1501
2001
2501
Pixel number
Page 24/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration: use of specific landscapes

Rayleigh scattering






For short wavelengths (B1, B2)
Specific viewing conditions (clear ocean, off-nadir viewing)
Use of B3 for aerosol optical thickness estimation
Use of meteorological data (water vapor, wind, pressure)
Less convenient that for Vegetation (specific acquisitions, clouds…)
Desert sites





Cross-calibration with either Polder, Vegetation or SPOT (SWIR)
Sites supposed stable
Similar viewing conditions
for the reference sensor
Correction for atmospheric
effects and spectral
sensitivity differences
Replaces the lamp for
SPOT5: high frequency
acquisitions are possible
Page 25/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration: use of specific landscapes

Vicarious calibration over test-sites


Simultaneous image acquisition and ground characterization of
reflectance and atmosphere
Cooperation with



University of Arizona (86-98)
White Sands test-site (NM)
French laboratories (LOA & LISE)
La Crau (France)
Recent achievement of an automatic
radiometer CIMEL station:





Continuous ground and atmosphere
characterization
Phone link transmission of the data
Enables calibration of any sensor,
each time it overpasses the site
Operational in La Crau
Others are planned
Page 26/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Radiometric Calibration
 Absolute Calibration:

Cross-calibration: simultaneous acquisitions



IQ mode
HRVIR or HRG / Vegetation
Synthesis


Use of all these methods
to match a sensitivity
curve
Discrepancy between
methods:
6% visible
8% SWIR
Page 27/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Spatial resolution
 Refocusing:
Bi-image method:







Simultaneous viewing of the same landscape with both instruments
Fixed focus for the reference camera
Focusing mechanism of the
1.2
other is moved
Determination of the position
1.1
-16.6
giving the highest ratio of the
Fourier transform of
1
corresponding images
First try on SPOT1 (1994)
0.9
Operationnally used for
SPOT4 and SPOT5
MTF ratio (HRG1/HRG2)

Defocus model
M easurement
Vertex
0.8
Use of the autotest device




Only for SPOT5
Periodic square target
No absolute measurement
Monitoring of focus over time
0.7
-28
-24
-20 -16
-12
-8
-4
0
4
8
12
Focusing mechanism position
Page 28/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Spatial resolution
 Modulation Transfer Function assessment:

Bi-image method for relative comparisons

Point Source target (since SPOT3): needs on-ground team

Edge target: natural (pb of edge quality) or artificial (SPOT5 THR)

Final synthesis => MTF @ Nyquist in both directions for each band
One of our Xenon lamps
SPOT5 THR image
of our Salon-de-Provence target
Page 29/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS
Summary
 Lessons learned:




Continuity needed between pre-flight and post-flight activities
Need for an operational Center in charge of all in-flight image quality
activities
Strong involvement (means, people)
Continuous improvement of our methods & means:



More accurate
More versatile
Easier to perform
 For the future: Pléiades = SPOT High Resolution Follow-on




On-board simplification (no calibration device, no on-board registration,
non-continuous detection lines, non-linear radiometric response, …) +
improvement of performances (resolution, geolocation accuracy…)
=> complexification of ground image processing & calibration
Geometry: new test sites with high resolution/accuracy GCPs
Radiometry: non-linear normalization => specific steered acquisitions,
new methods (histograms)
Resolution: Artificial Neural Networks (focus & MTF), bi-resolution
 Interest of sharing reference data over test-sites (cf HRS-SAP)
Page 30/30
International Workshop on Radiometric and Geometric Calibration, 2-5 Dec 2003, Gulfport, MS