www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 1 SEOM INSARAP: Sentinel-1 InSAR Performance Study with TOPS Data ESA-ESRIN Contract 4000110587/14/I-BG Interferometric TOPS Chain Description Pau Prats, Rolf Scheiber, Francesco De Zan DLR-HR, Microwaves and Radar Institute, Pau.Prats@dlr.de
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 2 TOPS InSAR Chain - Particularities of the TOPS signal - Azimuth-dependent Doppler centroid - Doppler variation larger than azimuth sampling frequency - Burst mode (synchronization required, burst-wise processing) - Critical steps - Offset computation for coregistration - Interpolation - Azimuth spectral filtering - Selected strategy - Geometric coregistration + global offset estimation - Valid for stationary scenarios (or scenarios with slow deformation rates, e.g., PSI)
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 3 TOPS Coregistration - Azimuth-variant Doppler centroid introduces strong requirements in terms of the required azimuth coregistration performance. φφ eeeeee tt aa = 2ππff DDDD (tt aa )Δtt - If not done accurately, azimuth phase ramps remain, hence introducing phase jumps between bursts. - Coregistration requirement for Sentinel-1: ~1 cm (~0.001 pixels) for a jump smaller than ±1.5º. - In the frame of a project with ESA, a technique based on spectral diversity (named enhanced spectral diversity, ESD) was developed, implemented and tested with TSX TOPS data.
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 4 TOPS InSAR Chain [1] - Main Workflow - Backgeocoding - Coregistration - Nominal from geometry [2] (interpolation) - Global offset (ESD in azimuth) - Interferogram generation - Spectral filtering (optional) Orbit DEM Backgeocoding Range and azimuth offsets Slant phase Master Slave Coregistration (for every burst) Constant offset estimation (for every burst) Range shift (for every burst) Enhanced Spectral Diversity (ESD) Coregistration Optimization potential! - TOPS specific processing - Blue-coloured blocks are TOPS-specific - Burst-wise processing - Debursting and mosaicking performed at the end (for interferogram generation) - Approach validated with more than 100 TOPS pairs of different sensors (TSX/RADARSAT- 2/Sentinel-1), including time series. Azimuth shift (for every burst) Spectral filtering Interferogram generation Coherence estimation Debursting and sub-swath mosaicking Coregistered SLCs Interferogram generation Filtered SLCs Interferometric products (burst-wise) Mosaicked interferometric products [1] P. Prats, R. Scheiber, L. Marotti, S. Wollstadt, A. Reigber, TOPS Interferometry with TerraSAR-X, IEEE Trans. on Geosci. and Remote Sens., vol. 50, no. 8, Aug. 2012. [2] E. Sansosti, P. Berardino, M. Manunta, F. Serafino, G. Fornaro, Geometrical SAR Image Registration, IEEE Trans. on Geosci. and Remote Sens., vol. 44, no. 10, Oct. 2006.
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 5 TOPS Azimuth Spectrum Discrete domain Continuous domain Folding! ESA document to compute the deramping function: Definition of the TOPS SLC deramping function for products generated by the S-1 IPF Reference: COPE-GSEG-EOPG-TN-14-0025 Ok!
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 6 TOPS Coregistration - Deramping and demodulation - Phase multiplication - 2D Interpolation - Offsets derived from geometry - Re-ramping and modulation - Interpolation of deramping and demodulation phases using the offsets - Complex phase multiplication Range and azimuth offsets Slave Deramping & demodulation 2D Interpolation Reramping & modulation Coregistered SLC
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 7 Enhanced Spectral Diversity (ESD) Master Geometric coregistered Slave - Exploitation of overlap eareas - Very efficient, since input data already available - Very robust (even for low coherence values) - Averaging of all overlap areas or polynomial fit - For a single slice (frame), averaging should suffice - For a combination of two or more slices, a polynomial fit might be better suited. - After estimation, shift each burst the given amount ( Azimuth Shift step) - Phase model at overlap areas might be extended to include baseline errors (similar as done in airborne). f a Read overlap regions of burst n and burst n+1 Generate upper and lower interferograms Generate differential interferogram and average Obtain phase and scale to pixels / Compute mean coherence of overlap region Last overlap region Yes No Obtain weighted mean value or polynomial fit t a Estimated azimuth shift(s)
www.dlr.de/hr > SEOM INSARAP INSARAP Workshop December 10, 2014 > Slide 8 Master Refine-coregistered Slave Azimuth Spectral Filtering - Needed in case of: - Limited burst synchronization accuracy - High Doppler centroid differences Burst mis-synchronization Master and slave Doppler centroids Burst mis-synchronization Doppler shift Crossing orbit Doppler shift Deramping with master deramping function Common band azimuth spectral filter Reramp with master deramping function BB cccccccccccc = BB aa ff mmmmmmmmmmmm DDDD ff ssssssssss DDDD + Δff sssssssss + Δff oooooooooo Non-parallel orbit Filtered master Filtered slave ff aa ff aa ff aa Deramping Filtering Reramping tt aa tt aa tt aa