1
VIS-SPEAR Calibration Update
Prepared forAir Force Research Laboratory
AFRL/VS
Aerodyne Research, Inc.
45 Manning Road
Billerica, MA 01821-3976
12 March 2002
2
Agenda
•VIS SPEAR Status – Calibration and Early Retrieval Examples Jones (45)
•VMP Calibration Procedures? Fetrow (15)
•MODTRAN 4-P Validation Status Conant (45)
•Feedback on the MODTRAN 4-P Validation Plan and Implementation Fetrow (15)
•Coordinated Measurements for VIS SPEAR and VMP at KAFB? All (15)
– Multiple Objectives
– Other Sensors: Research Scanning Polarimeter; AERONET Cimel
– Anticipated Schedule
•LWIR SPEAR Status Scott (15)
3
Meta-Agenda
• Recently achieved key SPEAR-TIP/PolTran milestones:– VIS-SPEAR is now working & field-worthy– VIS-SPEAR is spectro/polar/radiometrically calibrated– We are retrieving Stokes spectrum– Remote Surface Orientation Measurement (RSOM) patent
submitted– PolTran validation plan & results synergistic w/ NASA effort
• ~ $50K funded ARI effort remaining; additional $50K funded ARI set-aside for “field test”
Where-To From Here?
4
Where-To From Here?
• We ask both (1) for optimally directing remaining funds, and (2) to anticipate prospective efforts
• Possible Paths– VIS-SPEAR field collects -
• Targets; surfaces
• Atmospherics (IAW PolTran validation plan)– Likely NASA synergy
– PolTran aerosol parametric sensitivity determination– RSOM exploitation for target discrimination– (XX)IR-SPEAR?
Can we discuss (our roles in) AFRL SpectroPolarimetry Roadmap?
5
Recent & Prospective Progress
• We demonstrated calibrated “narrowband” (~14nm resolution) polarimetry against data.
• VIS-SPEAR is ready for local field data collects (FEB02)
• Steps remaining for VIS-SPEAR readiness for field test campaigns:– Improve mounting (CNC table too cumbersome)– Refine calibration; test against independent calibration
standards– Devise field-deployable calibration apparatus (existing lab
standard too cumbersome)– Integrate calibration/Stokes retrieval into data acquisition
program
6
Fringe Image
7
Fringe Spectra
Developed utility for real-time display of fringes to align sensor
8
SPEAR/AERONET Correlative Experiments
• Perform measurements with VIS-SPEAR and AERONET photometer collocated to assess radiometric and polarimetric performance of VIS-SPEAR.
• AERONET photometers have a long legacy (since 1993) and well planned calibration program.
• Experiment serves the interests of both AFRL and NASA.
9
Cimel Robotic Photometer
• Direct solar and sky radiance measurements.
• Sunrise to sunset.• Solar principal planes and
solar almucantars.• 8 pos. filter wheel w/ 4
interference filters. Additional 3 for polarization at 870 nm.
• Satellite feed and automatic data processing available via web.
• Rigorous calibration plan.
10
VIS-SPEAR Field Mounts
• Scans performed manually:– Almucantars: VIS-SPEAR mounted on
heavy duty tripod Alt-Az head w/ graduations.
– Principal Planes: VIS-SPEAR mounted on equatorial telescope mount adjusted for operation at 0 deg latitude.
• Scans performed automatically:– Full imagery: VIS-SPEAR mounted on
CNC rotary table. Very bulky – needs to be replaced for future field measurements.
11
Sensor ComparisonPrincipal Plane Scan
02/19/2002
0
5
10
15
20
-40 -20 0 20 40 60 80 100 120 140 160
View Angle [degrees]
Sky
Ra
dia
nce
[m
W/c
m^
2/sr
/um
]
CIMEL 675 nm 16:07UT
SPEAR 680 nm 16:35UT
12
VIS-SPEAR
Principal Plane Scan2/19/2002 16:33UT
0
1
2
3
4
5
6
7
8
9
10
-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90
View Zenith Angle [degrees]
Sky
Rad
ian
ce [
mW
/cm
^2/
sr/u
m] 480 nm BW=40 nm
520
560
600
640
680
720
760
800
820
13
Cimel Photometer
AERONET Sun Photometer02/19/2002
0
1
2
3
4
5
6
7
8
9
10
-10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170
View Angle [degrees]
Sky
Rad
ian
ce [
mW
/cm
^2/
sr/u
m]
1020 nm
675
441
870
14
SPEAR/CIMEL DoLPPrincipal Plane Scan
2/19/200216:30UT
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90
View Zenith Angle [degrees]
De
gre
e o
f L
ine
ar P
ola
riza
tio
n
CIMEL 870 nm
SPEAR 520 nm
SPEAR 680 nm
SPEAR 800 nm
15
VIS-SPEAR DoLPPrincipal Plane Scan
VIS-SPEAR2/19/2002
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90
View Zenith Angle [degrees]
Do
LP
480 nm BW=40 nm
520
560
600
640
680
720
760
800
820
16
AERONET Polarimetry
AERONET Billerica Raw Data 18 Feb 2002
0.1
1
10
-90 -60 -30 0 30 60 90
View Zenith (deg)
Radi
ance
Pol Channel 1 Pol Channel 2 Pol Channel 3
Billerica 18 Feb 2002
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
-90 -60 -30 0 30 60 90
View Zenith (deg)
Do
LP
16:10:33
17
Retrieval Context
• Retrieve Stokes spectrum from x(i)– x(i) is output of i-th pixel– Sensor input Stokes vector S().– Gaussian spectral blur h(), gain R(), and offset C(i).– Mueller matrix M; P-SIM crystals (lengths and birefringence of
material). Integration “over blur” – x(i) includes tails of x(i-1) and x(i-1).
• Retrieval first requires estimation of system parameters
)()())(,,()()()( 21 iCdSnMhRix
18
Calibration Sequence
• Cal Sequence - designed to estimate parameters of the system equation
– Spectral cal: • Mapping between wavelength vs. pixel index;
• Spectral blur h().
– NUC (Non-Uniformity Correction):• R and C
– Retardance cal: • L1, L2, n• and any additional system retardance
)()())(,,()()()( 21 iCdSnMhRix
19
Spectral Cal• Infer from spectral line source images:
– wavelength vs. pixel index; blur {h[] and }
• Hg(Ar) and Kr vapor lamps– line widths << pixel bin
• Non-linear regression using multiple lines:– retrieves: quadratic polynomial mapping between
and pixel index I; and (gaussian blur)
NLR retrieval matched the Andor and slit blur estimates
pixel bin: 2.3nm
0 50 100 150 200
5000
10000
15000
20000
20
Non-Uniformity Correction (NUC)
• NUC accounts for:– Spatial non-uniformity of pixel gain & offset
– Equalization of aperture shading (Cos4), spectral transmittance
– Maps raw digital counts to radiance units
• Requires spectrally-calibrated sources– Unlike conventional 2D imaging flat-fielding
Sli
t
500 1000 1500 2000Nanometers
0
50
100
150
200
250
300
A.U.
Cal source spectrum
0 50 100 150 200 2500
50
100
150
200
250
21
Retardance Cal
• Global non-linear regression retrieves from known input Stokes spectra (“Charlie Brown” frames):– crystal lengths L1, L2, quadratic polynomial n – nuisance parameters (for the calibration itself):
• i0 (intensity scale) • raop (unknown polarizer alignment offset angle)
• Employs Labsphere source (known spectral intensity) and rotating polarizer– absolute polarizer rotation alignment presently
unknown - known relative rotations between frames– assumes perfect, spectrally flat polarizer
22
Retardance Cal
550 600 650 700
0.0089
0.0091
0.0092
0.0093
Nominal vs. fitted polynomial n
900000950000
1´106
1.05´106
1.1´1061.9´106
1.95´106
2´106
2.05´1062.1´106
0100000200000300000400000
900000950000
1´106
1.05´106
1.1´106
• The objective (error) function for global NLR: – sum of squared errors:– error=data vs. parametric
signal model– Model parameters are L1,
L2, n, i0, raopError surface vs. crystal length
parameters L1 and L2
Highly non-smooth error surface requires Global NLR
23
Retardance Cal
0 20 40 60 80 1000
200000
400000
600000
800000
1´106
1.2´106
1.4´106
0 20 40 60 80 1000
250000
500000
750000
1´106
1.25´106
1.5´106
• Excellent model fit (calibration), except at short wavelength end for sideband-rich AoPs
• Fit error consistent with NUC, other errors ~1-2% (for now)A
oP=
6deg
AoP
=13
6deg
DataModel
24
Stokes Retrieval Strategy
• Invert the system equation– linear (pseudoinverse-”PI”) for Stokes parameters
• assumes spectrally-constant source across – non-linear regression (“NLR”) for:
• spectrally-constant {I,dop,aop,} across • parametric spectral source models (Sellmeier, Lorentz)• requires global solver
2
2 1
2 1
1
2
212 2
2 [ ]2
2 [ ]2 2
T
L nCos I
Q I dop Cos aopY L n L n
U I dop Sin aop CosSin Sin
V I dop Sin aop SinL n L n
Sin Cos
25
Stokes Retrieval to Date
• 6-sample (~14nm) Stokes retrieval achieved from PseudoInverse and NLR– Assuming flat source spectum
over 14nm band– But works nonetheless against
blackbody-like spectrum
• Handling spectral structure:– Will try NLR against spectral
models– Actual implementation: basis
pursuit?
0 20 40 60 80 1000
200000
400000
600000
800000
1´106
1.2´106
1.4´106
0 20 40 60 80 1000
250000
500000
750000
1´106
1.25´106
1.5´106
AoP
=6d
egA
oP=
136d
eg
DataModel