Comprehensive Vicarious Calibration and Characterization of a Small Satellite Constellation Using the Specular Array Calibration (SPARC) Method

Transcription

1 This document does not contain technology or Technical Data controlled under either the U.S. International Traffic in Arms Regulations or the U.S. Export Administration Regulations. Comprehensive Vicarious Calibration and Characterization of a Small Satellite Constellation Using the Specular Array Calibration (SPARC) Method 31st Annual AIAA/USU Conference on Small Satellites Logan, UT, USA Utah State University Tuesday, August 8, 2017 Session VI: Science Missions Payloads 1 Raytheon Intelligence, Information & Services Stephen J Schiller Marcus Teter Raytheon Space and Airborne Systems John Silny Copyright 2017 Raytheon Company. All rights reserved.

2 Introduction: Importance of Calibration Image data is more than just a Pretty Picture. Conversion from digital number (DN) to absolute units (i.e. radiance) allows physics-based exploitation for creating value added products and enhanced data mining. Exploitation capabilities with calibrated spectral data include: Searching pixel data for specific spectral signatures (target detection), Finding changes at a geographic location between multiple scenes (change detection), Assigning a label to each pixel surveying scene content (classification), Apply atmospheric corrections to derive surface reflectance and BRDF Vicarious calibration monitors sensor calibratebility affected by stress of launch, harshness of space, and degradation over the lifetime of the mission. For a constellation, calibration is necessary for the data and data products to become sensor independent. Calibration is essential. On-board calibrators are not when using vicarious methods.

3 Mainstream Vicarious Calibration Methods are Intended to Verify Prelaunch or On-board Derived Absolute Gains But Not Much More Large area instrumented or pseudo-invariant desert sites are used to provide a known at-sensor radiance to verify sensor radiometric performance and derive knowledge of biases between sensors. Each site requires a calibration model to account for seasonal changes in solar illumination, BRDF reflectance and atmospheric conditions with interpolation to the resolution of the sensor under calibration. Railroad Valley, Ivenpah, Tuz Galu, Algeria 4 Only one radiance level is provided at each target (almost all are bright) The vicarious calibration targets are assumed large enough that the system spatial resolution does not affect the vicarious derived gain coefficients. radiometric gain coefficients from these sites do not apply to small targets. (< 4x FWHM of the system PSF, pixels) Copyright 2017 Raytheon Company All Rights Reserved

4 The SPARC Method Offers a Innovative Calibration and Image Assessment Capability Using Vicarious Reference Targets An optimal calibration capability should not only deliver radiometric performance analysis but support a complete image assessment system Method can be incorporated as part of a ground station, data production and archiving architecture Key assessment attributes include: Radiometric validation, updates, uncertainties and traceability Spatial PSF and MTF characterization (Along-track & cross track differences) Image quality analysis (Focus, jitter, data compression, focal plain uniformity) Geometric calibration (camera model) Band-to-band registration Spectral band ratio trending and inter-sensor band difference coefficients Validate a sensor s Calibration Parameter File (CPF) for each satellite. Monitor inter-sensor performance for transformation to a fleet average radiometric scale. Build databases to support detailed spatial, geometric and radiometric bulk trending for monitoring sensor degradation over the lifetime of each sensor. 8/10/ Copyright 2017 Raytheon Company All Rights Reserved

5 Specular Array Calibration (SPARC) System: Technique is based on generating a solar star as a calibration source from a spherical mirror. An array of spherical mirrors create points sources all with an identical solar spectrum. Reflected intensity of targets are constant for all view geometries (No BRDF effects). Targets are small and easy to deploy. Typically much smaller than the sensor Ground Sample Distance (GSD) Provides radiometric calibration, spatial, spectral, and geometric calibration references. SPARC makes available a state-of-the-art solution for post launch sensor calibration, performance optimization, and quality assessment to data suppliers and users across the industry. Copyright 2017 Raytheon Company All Rights Reserved

6 Conceptualizing The SPARC Method The SPARC method allows any earth observing sensor to be calibrated to the solar spectral constant just like a solar radiometer. The reflective Field-of-Regard for a mirror acts as a Field-of-View (FOV) aperture stop allowing the sun to be viewed directly as an absolute reference. Imaging Sensor Sun FOV Point Source Array Reflected Field of - Regard Aperture Stop SPARC Mirror FOV Aperture Stop Sun Photometer Radiometric Panel Detector Solar Image By selecting an appropriate radius of curvature of the mirror, it scales down the brightness of the sun to an intensity that does not saturate the sensor s focal plane and does so without changing the upwelling spectral properties. Copyright 2017 Raytheon Company All Rights Reserved

7 SPARC Targets Isolate the Direct Solar Signal The SPARC reflectors act as a spectrally flat neutral density filter allowing the sensor to look directly at the sun through the same atmosphere as the rest of the scene. Two sequential IKONOS images recorded on the same overpass. Total at-sensor radiance of each target is quantized by the number of mirrors. When the sensor moves outside the mirror s field-of-regard, the images show how the direct solar component turns off showing the methods ability to subtract out the background and atmospheric radiance pattern and isolate the direct solar signal.

8 SPARC Target Point Sources Support Detailed Analysis of the Full 2-D PSF Profile SPARC uses a grid of spherical reflectors to create an oversampled point spread function (PSF). IKONOS Image Of point targets Extracted images have different pixel phasing After centroiding, images are combined to reveal oversampled 2-D PSF Profile 2-D PSF based on images taken two years apart show similar asymmetric profile Level 0 Level 1 (Resampled) Level 1 + MTFC Analysis evaluates the image quality at each step along the image processing chain. 8/10/2017 8

9 Radiometric Calibration Methodology SPARC Predicted At-Sensor Radiance/Mirror sensor and collection geometry specific - Watts/(m 2 sr micron)/mirror L ( λ, θ ) at sensor r = ρ( λ, θ ) τ ( λ) τ ( λ) E ρ (λ,θ r ) = Mirror specular reflectance at the reflectance angle θ r τ (λ) = Sun to ground transmittance τ (λ) = Ground to sensor transmittance E o (λ) = Solar spectral constant R = Mirror radius of curvature (m) GSD = Line-of-site ground sample distance (m), cross-scan (x) and along-scan (y) r o 2 R ( λ) 4GSD( x) GSD( y) Integrated Digital Number (DN) Response/Mirror SPARC target DN response measured from image 9 Integrated Response = DNint = DN( n) DN background n= 1 [ ] Slope of regression line gives response/ mirror Spectral Band Slope: DN/Mirror R 2 Blue Green Red NIR Sensor Absolute Gain Response = (ΣDDDD iiiiii (λλ)/mmmmmmmmmmmm) (LL aaaa ssssssssssss (λλ,θθ rr )/mmmmmmmmmmmm) Validated at < 3% Pan and < 2.5% MSI reproducibility per overpass 8/10/2017 9

10 SPARC Allows Landsat and Sentinel 2 to Become A Radiometric Member of the Small Sat Constellation Landsat 8 Pan image of SPARC targets PICS cal targets vary with GSD, SPARC does not! Landsat will always be maintained as the Gold Standard for VNIR/SWIR radiometric calibration Applying the same calibration methodology to Landsat, Sentinel 2, and each constellation member places all on the same traceable radiometric scale. The higher spatial resolution, temporal frequency and variable view geometries offered by a small sat constellation intercalibrated with Landsat, can create an effective disaggregated Landsat fleet with the potential of achieving improved exploitability and new product development.

11 Summary The SPARC method provides an robust and reproducible vicarious calibration method ideal for use with small sat constellations SPARC imparts confidence to the user community in the performance of the sensor and the quality of the data products. For a constellation all sensors are transformed to a common fleet average radiometric scale. Analysis isn t done on a thousand independent collects, but done on a single data base containing information from a 1000 collects. SPARC enhances small sat data so that it can be aggregated into Big Data 8/10/

12 Evaluating Your Data Quality with SPARC Method Open invitation to data providers and data users to participate in Oct 2-6, 2017 SPARC Targets will be deployed at the Raytheon, El Segundo South Campus for imaging by sensors with GSD of 6m or less Lat. = 33 o N, Long. = 118 o W Contacts: Stephen Schiller John Coogan Stephen_J_Schiller@Raytheon.com John_J_Coogan@Raytheon.com (310) (972) /10/

13 Back-up Charts 8/10/

14 Calibration with Small Diffuse Reference Targets are Affected by Sensor Spatial Performance Contrast reduction by the sensor system MTF introduces an error in the derived sensor gain if one applies the reflectance-based vicarious method using small targets. Even though the predicted at-sensor radiance values may be accurate, the relative contrast loss affects the measured image DN values used to derive the gain coefficient introducing and unknown scaling error. The sensor MTF makes bright targets fainter and dark targets brighter relative to the average background radiance The result is a gain value (slope = DN/radiance) that is too low compared to the true gain. Small target radiometry requires knowledge of both radiometric response and spatial image quality In scene Lambertian targets used for reflectance/radiance calibration DN values increased by contrast loss Central 2x2 pixel DN values are used to estimate response to the at-senor radiance from each target. Effect of MTF on Small Target Reflectance-Based Vicarious Calibration Gain Measurement Truth (MTF = 1) Measured (MTF < 1) Slope decrease due to MTF contrast reduction Copyright 2017 Raytheon Company All Rights Reserved DN values decreased by contrast loss 14