Photo by Sassan Saatchi UAVSAR in Africa Quality Assurance and Preliminary Results Brian Hawkins, UAVSAR Team CEOS SAR Cal/Val Workshop 2016 Copyright 2016 California Institute of Technology. Government sponsorship acknowledged.
Overview UAVSAR Instrument AfriSAR campaign Onboard processor Quicklook processor Quality Assurance checks Initial results 2
UAVSAR Instrument Uninhabited Aerial Vehicle Synthetic Aperture Radar Precision autopilot can repeat trajectories within a 10 m tube. Electronically scanned antenna for Doppler control. Frequency Bandwidth Resolution Polarization Polarization Isolation Radiometric Calibration NESZ Nominal Altitude Swath Width 1.26 GHz (L-band) 80 MHz 2.5 m Range, 1 m Azimuth Full Quad-Polarization < -30 db 0.7 db -30 to -50 db 12.5 km > 20 km Reconfigurable P-band antenna Ka-band interferometer Global Hawk UAV A. G. Fore et al., "UAVSAR Polarimetric Calibration," in IEEE Transactions on Geoscience and Remote Sensing, vol. 53, no. 6, pp. 3481-3491, June 2015. 3
AfriSAR Campaign Gabon offers large areas of pristine tropical forest, important for understanding of the global carbon cycle. Acquire SAR, LIDAR, and in situ measurements of three dimensional forest structure. Collaboration including ESA, DLR, NASA, and AGEOS (Gabon) supporting remote sensing missions (BIOMASS, NISAR). Operating far from home means early detection of anomalies is crucial. 4
AfriSAR Campaign Example UAVSAR flight plan. 14,000 km from JPL! 5
Onboard Processor Allows evaluation of fully focused imagery in real time. Implemented in hybrid microprocessor + FPGA (Virtex 5) architecture. Radiation tolerant and low power for potential space applications Patches are geocoded and converted to KMZ format for display in Google Earth. Some limitations: reduced swath, single polarimetric channel. 6
Onboard Processor Packed, formatted data Second channel to additional OBP board Focused data, to product generation Microprocessor Board PMC Interface Board Data Unpacking Header Stripping Headers (time tags and motion data) Real-time Header Processing Kalman filtering of motion data Radar Timing Ring Buffer Platform Timing Ring Buffer Radar data At start-up Load V5 FPGA bitfiles Sinc interpolation coeff. Chirp for range comp. Range Compression Parameters FIFO Buffer Presum Buffer (2GB) FPGA Processor Board Presum Motion Compensation Corner Turn Buffer (2GB) FFT/Range Migration Correction Parameters FIFO Buffer Azimuth Compression Calculate OBP processing parameters Ring Buffer (Range) Ring Buffer (Azimuth) Range Processing FPGA PCI Interface FPGA SelectMap Bus (bit file load) Azimuth Processing FPGA cpci Bus Y. Lou et al, "Onboard Radar Processor Development for Rapid Response to Natural Hazards," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 9, no. 6, pp. 2770-2776, June 2016. 7
Onboard Processor Real-time processed image of Lope National Park, Gabon. Acquired March 8, 2016. 8
Quicklook Processor After flight we transfer the raw data to other storage media and perform sanity checks. Quicklook processor uses simpler algorithm than OBP but can produce imagery for full swath and all channels quickly on a laptop. No motion compensation Fixed Doppler No range cell migration (RCM) correction Limited azimuth resolution (chosen so RCM < ¼ Δr) Completes in a few minutes (I/O bound) 9
Quicklook Processor Range Azimuth Lope National Park (same image) HH HV VV 10
Quicklook Processor Range Azimuth Pongara, Gabon Feb 27, 2016 HH HV VV 11
Quicklook Processor HH Pongara, Gabon March 8, 2016 Trouble in HH! Poor focus and reduced contrast. VV Need more information 12
Quality Assurance Checks Also ported automated quality assurance tools to laptop system. Check lots of raw data metrics Gain setting / saturation Radar PRF / duty cycle Raw data spectrum Noise spectrum Internal calibration signals Automated pass/fail criteria 13
Quality Assurance Checks Internal calibration signals Active Phased Array Antenna TR Module RX _H ADC1 RFES RX1 RX1 Cable RXH 24-WAY TX ANT _H AWG ADC1 A K RFES Exciter RFES RX2 TX Cal RX Cal Cal Tone RFES Switch Network TX Cable RX2 Cable B J Antenna Switch Network C G TX 24-WAY RXV 24-WAY CAL 24-WAY D F RX_V E ANT _V O Chirps are sent through TX and RX paths at about 1 Hz Tone injected into every received pulse Can evaluate phase/timing stability 14
Quality Assurance Checks QA tool flags problems on H-transmit and generates plots. Intermittent problem with waveform generator. Was able to re-plan flights to reacquire data. Would not have known without near-real-time analysis tools. 15
Initial Results Acquired interesting data over mangroves and tropical forests. Multiple geometric and temporal baselines for forest structure investigations. Evaluating TomoSAR processing mode Coregistration and residual motion corrections for non-zero spatial baselines. For now PolSAR and repeat-pass data are available. http://uavsar./ 16
Initial Results PolSAR (Mondah_27080) 17
Initial Results InSAR Stack (Mondah_27080) 18
Acknowledgement UAVSAR Team Especially Kean Tham and Naiara Pinto, who operated UAVSAR in Gabon. 19