1. Headwall Instrument Overview Laboratory Characterizations Geo-Location Field Characterization Data Product Description References

2. 2 Headwall Hyperspectral Imager Background: COTS Pushbroom Imaging Spectrometer (Concentric Offner) Acquired for demonstrations on Ikhana UAV Not designed for quantitative measurements Description: Spectral Range: 350.8nm to 863.8nm Spectral Range (useful): 400 – 800nm (no OSF) Bands: 172 (43 @ 4x binning)  nm  12 nm @ 4x binning) Spatial Pixels: 400 (4x binning) IFOV/FOV: 2.33 mrad / 49.9 degrees (4x binning) Array: 1600 x 1200 pixels (14.8  m) @ -30C Grating optimized at 450 nm Spectral Smile: <3 nm Keystone: <3 nm Airborne Sensor Facility Ames Research Center

3. 3 Airborne Sensor Facility Ames Research Center Camera Head, IMU, Mounting Frame & Base Plate Twin Otter Installation

4. Laboratory Characterizations Radiometric Calibration The instrument was mounted on a Goniometer over the Ames 30” integrating sphere. The sphere is NIST-traceable, with ~3% uniformity across the aperture. The sphere was viewed at three angles (0 and +/- 7 deg.), over 6 lamp levels (0,1,3,6,9,12 lamps) An intercept term, and linear radiance equations were generated on a pixel by pixel basis.

5. Focal Plane Array j (Wavelength) i (Spatial) j (Wavelength)

6. This array represents one line i (Spatial) j (Wavelength) For cell (i,j) in the array, Radiance i,j = m ij (DN L,i,j -DN 0,i,j )+b ij DN L = digital number at radiance L DN 0 = digital number with shutter closed

7. Compute slope and intercept: (m, b) i,j Pseudo code: DN 0 = 800x600 average of shutter closed frames (grab 0-10, grab (N-10)-N ) m, b = 800x600 arrays for (i=0,…,800) for (j=0,…,600) for L k, k={1,2,3,4,5,6,9,12,…} lamps DN L,i,j = average of cell (i,j) at radiance L k SD L,i,j = SD of cell (i,j) at radiance L k if, DN L,i,j >= 4094 – 2* SD L,i,j, throw out this data point end fit L i,j = m ij (DN i,j -DN 0,i,j )+b ij by least squares. set m ij, b ij end

8. Assumptions The response is linear for radiance L k is uniform across the spatial dimension The sphere output is characterized well, spectrally and radiometrically DN L,i,j >= 4094 – 2* SD L,i,j detects saturation ??

9. dark

10. 6 lamp array

11. Instrument Response at 3 Look angles over the Sphere

13. Signal-to-Noise Ratios over Ames Sphere (at the wavelengths where the sphere signal approximates the ocean signal)

14. Laboratory Characterizations Spectral calibration Instrument mounted on a Goniometer over a 10” collimator, illuminated with a NIST-traceable Hg line emission lamp Source viewed at +/-7, +/-14 and +/-21 degrees view angles Hyperspec software took a "base" wavelength value - the center of the lowest wavelength pixel - and slope term to make a linear fit to the Hg lamp peaks

15. Hg Lamp Lines (with zoom for spectral smile & keystone assessment)

16. Laboratory Characterizations Polarization Sensitivity

17. lens TE cooled camera integrating sphere scrambler polarizer spectrometer

18. Measurements taken –Lens only –Lens and polarizer at 9 o’clock –Lens and polarizer at 12 o’clock –Lens, scrambler, and polarizer at 9 o’clock –Lens, scrambler, and polarizer at 12 o’clock –Scrambler and lens –Dark fields were recorded for all measurements

19. no scrambler, 12 o’clock / 9 o’clockwith scrambler, 12 o’clock / 9 o’clock

20. Geo-Location The geo-referencing is based on attitude and position data measured by the Applanix POS-AV and GPS at the precise moment that a frame is captured by the camera. Unfortunately, the OKSI software does not enforce the synchronization of events sent to the POS and frame capture by the camera Unaccounted dropped frames and drift between the camera frame time and the POS event time affect the accuracy of the geo-referencing Lines without additional processing (including all ocean lines) are estimated to have accuracy between 1-6 pixels (7-42 meters @ 10,000 ft AGL)

21. Field Characterizations

22. Monterey Test Flight: Headwall Retrieval vs. MODTRAN Predicted Radiances (Clinton, 2011)

27. “L-1B” Data Format Notes: All files are binary format, BIL, with ENVI headers.  spectral bands (  nm) Recommend processing to 4x binning (  nm) Spatial Pixels: 400 (4x binning done in collection) *.cub files are the raw data as written by the Headwall software. Units are digital number. *_rad files are calibrated at-sensor radiance. Units are 100* W/m 2 /sr/nm. *_igm are image geometry model files suitable for geo-referencing.cub or _rad files in ENVI. *_geo files are 3-band (RGB, or NIR-R-G false color composite), geo-referenced files, which serve as “quick looks” and are produced using default values for pixel size, re-sampling, projection, etc. *_glt files are ENVI geographic lookup files used to produce the _geo files. *.house are ASCII formatted POS output of sensor position and attitude for line in the imagery. The events are based on either original time or derived time depending on flight line.

28. Nick Clinton’s 2011 Headwall COAST Blog: https://teams.share.nasa.gov/arc/r/astlcollaboration/headwall/default.aspx Poster on Characterization of Headwall Spectral Distortion Available on request References