Radiometric Calibration

1
Radiometric Calibration

• Stuart F. Biggar, U of AZ
• Kurtis J. Thome, U of AZ
• Simon J. Hook, JPL

2
Outline

• Preflight Biggar
• Characterization
• Calibration
• On-board calibration systems
• Sensor artifacts
• In-flight VNIR and SWIR Thome
• In-flight TIR - Hook

3
Calibration (absolute)

• Why do we calibrate?
• Understand sensor performance
• L1 data in physical units (at sensor)
• Radiance in Watts/m2-sr-micrometer (or similar)
• Determine sensor linearity
• Derived units
• Reflectance
• Geophysical units (temperature, etc)
• Comparison to other sensors
• Atmospheric correction
• Reasonable looking imagery

4
Calibration (2)

• What do we need to be able to calibrate?
• Stable instrument
• Over time
• With changing environment
• Characterized instrument
• Spectral response
• Spatial response
• Noise
• Systematic errors
• Image artifacts
• Stray light
• Stable calibration source

5
Characterization

• Piece part measurements
• Filter transmittance
• Lens and/or mirror transmittance and reflectance
• Mirror reflectance and scatter (BRDF)
• Lamp output, stability, longevity
• Detector Quantum efficiency, spectral response,
  noise, temperature dependence

6
Filter spectral characteristics
7
Normalized Response
8
Characterization (2)

• Assembly level
• Alignment
• Relative spectral response (of telescope)
• Noise amplifier assembly for example
• Field-of-view
• Stray light

9
Characterization (3)

• Sensor level
• Relative spectral response
• Field-of-view
• Modulation transfer function
• Noise
• Random
• Coherent
• Stray light

10
ASTER design features (1)

• Long linear array detectors with long strip
  filters in VNIR and SWIR pushbroom
• Good SNR due to longer integration time
• Detector and filter uniformity issues
• Many detectors (5000 in VNIR for example)
• Gain
• Offset
• Spectral response
• Large focal plane
• Post detector electronic chains (amps, A/D, etc)
• Stripes in images even of constant radiance
  scenes

11
ASTER L1A,bands 1and 13
12
ASTER design features (2)

• Short linear arrays (10 elements) with a scan
  mirror in the TIR sensor whiskbroom
• Shorter integration time
• Larger pixels to improve the SNR
• Limited number of detectors
• Individual amplifiers
• Stripes in images but repeated along track

13
ASTER design features (3)

• Three separate sensors
• 3 telescopes
• Individually pointed
• Alignment
• Rotation of images
• Focal plane distortions
• 3 manufacturers
• Different design practices
• Different measurement practices
• Different calibration methodologies and equipment

14
(No Transcript)
15
(No Transcript)
16

17
Radiometric calibration

• Preflight in the laboratory
• Well characterized and controlled source
• Reasonable but not normal environment
• May (or may not) approximate normal imaging
  operation mode
• Scan mirror may not be operating
• Source is usually smaller angular extent than the
  earth
• Uniform source is not a normal image

18
On-board calibrators

• VNIR
• Lamp based source with monitors
• SWIR
• Lamp based source with monitors
• TIR
• Blackbody that can be heated

19
Philosophy

• Careful preflight calibration
• Determine absolute response so we can convert
  from DN to radiance
• At the same time, run the on-board systems to
  transfer a radiance value to them using the
  sensor (ASTER)
• Determine noise (random and coherent)
• Look for image artifacts
• Use preflight values after launch
• Update after launch as sensors usually change
• On-board
• Vicarious (ground reference and moon)

20
Preflight source for VNIR/SWIR

• Large aperture integrating sphere
• 1 meter diameter sphere
• Round output port
• Multiple lamps
• Internal surrounding the port
• External through small ports
• High reflectance coating on the sphere wall
• Barium sulfate (BaSO4)
• Spectralon (sintered PTFE)
• Gold sandpaper (rough, diffuse surface)

21
40 BaSO4 integrating sphere
22
Integrating sphere

• Advantages
• Uniform radiance across port (one or so)
• Lambertian (radiance independent of direction)
• Multiple output levels
• Change number of lamps (MELCO)
• Change voltage/current of all lamps (voltage for
  NEC sphere)
• Change aperture size between external lamp and
  sphere
• Reasonably stable with good control of lamp (V or
  I)
• Repeatable if temperature is controlled
• Disadvantages
• Low output in blue relative to NIR and SWIR

23
TIR preflight cal source

• Large area blackbody
• Measure the temperature precisely and accurately
• Measure (or compute) emissivity
• Compute radiance output
• Operated in a vacuum chamber with the sensor
• Vary temperature of source to get multiple
  radiance values
• Absolute calibration
• Linearity correction (output usually fit to
  something other than a gain and offset
  quadratic is common)

24
Calibration chain

• NRLM (Japanese equivalent of NIST) calibrated a
  set of fixed point blackbody simulators
• Calibration transferred to portable, variable
  temperature blackbody simulators
• Calibration transferred to the ASTER calibration
  sources (Spheres and TIR calibration blackbody)
• VNIR and SWIR spheres were measured at NEC and
  MELCO with a set of transfer radiometers from
  NIST, NRLM, GSFC, and University of Arizona
• NRLM (now AIST) and NIST collaborate to ensure
  that their scales are consistent

25
Are preflight values usable?

• Sometimes
• Launch may cause a shift in a sensor performance
• Operating temperatures may be different
• Something may have moved
• On-board lamps may change output
• Convection inside the lamp may be different
• Aging of the lamp
• Emissivity of the on-board blackbody may change
• Mirror reflectance may change
• Filters/detectors/amplifiers may change
• There may have been unexpected features

26
Unexpected features

• VNIR
• On-board calibrator monitor is off-scale
• Change in output is more than expected so dynamic
  range of A/D is too small
• SWIR
• Unexpected stray light causes crosstalk
• Present during preflight calibration and in all
  in-flight data

27
VNIR OBC

• VNIR OBC has two monitors for each of two
  redundant calibrators
• One monitor diode at lamp
• One monitor diode at output
• They track but at different rates
• We really want the output monitor to determine
  how well the on-board calibrator is working but
  we have only the lamp monitor on-scale

28
VNIR OBC output monitor

• NEC selected an expected output range based on
  preflight measurements, OPS experience with a
  similar calibrator, and desire to maximize
  resolution
• Output has fallen to below the lowest expected
  monitor output
• Telemetry value for monitor is now a flat line
  (signal offset lt zero)

29
SWIR crosstalk

• SWIR has multiple spectral bands
• 6 linear PtSi detector arrays
• Spectral selection filters over the arrays
• Not all light hitting the detector is absorbed
• Light hitting between the detectors is reflected
• Some reflected light is reflected back down by
  the filters
• Optical crosstalk

30
(No Transcript)
31
(No Transcript)
32
Stretched RGB of a Japanese Island, SWIR bands
4,5, 9, 400x400 pixels
33
SWIR Crosstalk

• Present in all images
• Preflight calibration
• In-flight calibration
• normal images
• Visible in images with strong contrast
• Not visible but present in others
• Currently NOT corrected for in normal processing
  of L1 data

34
Crosstalk correction

• If we know the amount of light leaking from one
  pixel to each other pixel in all the SWIR bands,
  we can correct for it
• Preflight data for MTF determination would
  probably contain much of the needed information
  but it was recorded only for the band under test
• Scan a line source across the array in both
  directions
• Scan a point (pixel size or less) in both
  directions
• It is possible to infer the correction from
  images with strong contrast (coast lines,
  islands, moon, and similar), however it is
  difficult and incomplete

35
SWIR correction

• Japanese team has developed a beta level,
  Windows (Win32) based, correction program. It
  does one scene at a time operating on a L1B HDF
  input file and writing a corrected L1B HDF file.
• US team is developing a program that is run as
  part of L2 processing. It starts with L1A data.
  For example, you will be able to order
  atmospherically-corrected, crosstalk-corrected
  surface reflectance at L2.

36
Corrected image, SWIR RGB with 4,5, 9
37
SWIR Crosstalk Correction

• Qualitatively improves image
• Largest effect is band 4 into 5 and 9
• However, any band should leak into all others
  (including itself) with the strongest effect on
  adjacent bands on the focal plane
• Band 4 has higher typical radiance than the
  others
• Japan correction has only band 4 into others
• There is interplay between crosstalk and water
  vapor absorption, especially in band 9

38
Conclusions (preflight)

• ASTER was calibrated preflight
• VNIR accuracy was probably within spec
• SWIR accuracy is poorer due to crosstalk
• TIR was probably within spec
• Preflight calibration is probably not appropriate
  at this point
• Change in sensor (VNIR and some TIR bands)
• Crosstalk in SWIR
• Calibration is being updated