S1TBX-SE-10 Joint coregistration approach for large TOPS-stacks
Description
TOPS coregistration requires an accuracy of 1 millipixel. The standard approach is to first coregister images based on a DEM, and subsequently refine the shifts using incoherent cross correlations and spectral methods, such as enhanced spectral diversity (ESD). The current version of SNAP/S1TBX provides support for azimuth offset estimation based on ESD. However, the method is applied w.r.t. a single master image, which is likely to fail for long stacks (usually highly affected by temporal decorrelation). This enhancement will apply the ESD method to all possible interferometric pairs (master-slave and slave-slave combinations). Note that, unlike the multi-master network defined for S1TBX-SE-2, the ESD network only operates on TOPS burst overlaps, which reduces the computational cost and allows for automating the process with minimal user interaction. For example, the user could specify a maximum temporal baseline and let the algorithm decide which remaining pairs to take based on coherence.
Intermediate (internal) deliverables (this planning may be adjusted):
Sprint 1 (done):
Improve ESD operator so that it uses a periodogram-based estimation of the azimuth shift (see "Coregistration of Interferometric Stacks of Sentinel-1 TOPS Data, EEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 14, NO. 7, JULY 2017").
Sprint 2 (done):
Test and demonstrate improvement of periodogram-based estimator.
New workflow design for joint coregistration.
Sprint 3 (done):
Further test periodogram-based estimator in selected areas (according to users' reports on the forum).
Finish first version of joint coregistration (something that works, even if the workflow is not yet ideal).
Sprint 4 (done):
Improve workflow.
Update standard processing graphs.
Sprint 5 (done):
More test results (e.g., different areas).
Improve performance.
Improve documentation.
Environment
None
GitHub Work
None
Activity
Show:
esteban.aguilera
17 June 2020 at 19:28
@Lisa House@Luis Veci
You can find all the work related to this ticket here:
TOPS coregistration requires an accuracy of 1 millipixel. The standard approach is to first coregister images based on a DEM, and subsequently refine the shifts using incoherent cross correlations and spectral methods, such as enhanced spectral diversity (ESD). The current version of SNAP/S1TBX provides support for azimuth offset estimation based on ESD. However, the method is applied w.r.t. a single master image, which is likely to fail for long stacks (usually highly affected by temporal decorrelation). This enhancement will apply the ESD method to all possible interferometric pairs (master-slave and slave-slave combinations). Note that, unlike the multi-master network defined for S1TBX-SE-2, the ESD network only operates on TOPS burst overlaps, which reduces the computational cost and allows for automating the process with minimal user interaction. For example, the user could specify a maximum temporal baseline and let the algorithm decide which remaining pairs to take based on coherence.
Intermediate (internal) deliverables (this planning may be adjusted):
Sprint 1 (done):
Improve
ESD
operator so that it uses a periodogram-based estimation of the azimuth shift (see "Coregistration of Interferometric Stacksof Sentinel-1 TOPS Data, EEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 14, NO. 7, JULY 2017").
Sprint 2 (done):
Test and demonstrate improvement of periodogram-based estimator.
New workflow design for joint coregistration.
Sprint 3 (done):
Further test periodogram-based estimator in selected areas (according to users' reports on the forum).
Finish first version of joint coregistration (something that works, even if the workflow is not yet ideal).
Sprint 4 (done):
Improve workflow.
Update standard processing graphs.
Sprint 5 (done):
More test results (e.g., different areas).
Improve performance.
Improve documentation.