Gamma DIFF&GEO: Reference Manual

S1_coreg_overlap

NAME
S1_coreg_overlap - Script to determine a co-registration offset based on the burst overlap applying a "spectral diversity method".

SYNOPSIS
S1_coreg_overlap <RSLC1_tab> <RSLC2_tab> <pair> <off> <off_out> [cc_thresh] [fraction_thresh] [ph_stdev_thresh] [cleaning] [RSLC3_tab]

<<RSLC1_tab>>	(input) 3 column list of TOPS master image (SLC, SLC_par, TOPS_par; row order IW1, IW2, IW3)
<RSLC2_tab>	(input) 3 column list of TOPS slave image (SLC, SLC_par, TOPS_par; row order IW1, IW2, IW3)
<pair>	(input) ID used for InSAR (e.g. 20141003_20141015)
<off>	(input) offset parameter file (with refinement offset polynomials)
<off_out>	(output) corrected offset parameter file (with refinement offset polynomials)
[cc_thresh]	coherence threshold used (default = 0.8)
[fraction_thresh]	minimum valid fraction of unwrapped phase values used (default = 0.01)
[ph_stdev_thresh]	phase standard deviation threshold (default = 0.8 radian)
[cleaning]	flag to indicate if intermediate files are deleted (default=1 --> deleted, 0: not deleted)
[RSLC3_tab]	(input) 3 column list of already available co-registered TOPS slave image to use for overlap interferograms

EXAMPLE
S1_coreg_overlap 20141003.SLC_tab 20141015.RSLC_tab 20141003_20141015 20141003_20141015.off 20141003_20141015.off_out 0.7 0.02 0.8 1

Determines a co-registration azimuth offset refinement based on the burst overlap regions, applying a "spectral diversity method". The updated azimuth offset refinement is written out to the output offset parameter file (20141003_20141015.off_out).

DESCRIPTION
S1_coreg_overlap is used in the S1 IWS co-registration procedure or to assess the co-registration quality of co-registered S1 IWS SLCs.

As input a pair (master and slave) of two already almost perfectly co-registered Sentinel-1 (S1) Interferometric Wide Swath (IWS) burst SLC are provided. These burst SLCs consist of the sub-swath burst SLCs for one or several sub-swaths as defined in the SLC_tabs provided for the reference (e.g. 20141003.SLC_tab) and the slave (20141015.RSLC_tab). For each sub-swath the SLC_tab contains one line with the burst SLC name (e.g 20141015.IW1.rslc), the name of the SLC parameter file of the burst SLC (e.g 20141015.IW1.rslc.par), and the related TOPS parameter file (e.g. 20141015.IW1.rslc.TOPS_par). In this IW1 is used to indicate that it is for the first sub-swath (IW1) and rslc is used to express that it is an already co-registered slave burst SLC.

The indicated pair parameter (e.g. 20141003_20141015) is used in the naming of the intermediate files automatically generated by the program.

Then an input and an output offset parameter file need to be indicated. The input offset parameter file is the file that was used in the co-registration of the burst SLC pair (as used when resampling the slave burst SLC using the program SLC_interp_S1_TOPS). The input offset parameter file can already contain offset polynomials e.g. as determined in a refinement using intensity matching or in a previous iteration using S1_coreg_overlap. The output offset parameter file corresponds to a copy of the input offset parameter file with an update to the constant value of the azimuth offset polynomial.

To apply the refined azimuth offset refinement you need to rerun the slave burst SLC resampling, but using the output offset parameter file instead of the input offset parameter file (when running the the program SLC_interp_S1_TOPS).

S1_coreg_overlap applies a spectral diversity method to all the available burst overlap regions of all the available sub-swaths. For this it determines for each burst overlap regions which sections of the earlier and later burst are overlapping. For each overlap region these two sections are copied out of both the master and slave burst SLC, resulting in four small SLC sections of identical size. Single look interferograms are then calculated for the pair of SLC sections (master and slave) of the earlier burst, and for the pair of SLC sections (master and slave) of the later burst. Next the phase difference between these two single look interferograms is calculated to get the "double-difference" interferogram. This complex valued interferogram is then multi-looked and unwrapped. Before the unwrapping a coherence measure is calculated and used as weighting function in the minimum-cost-flow unwrapping (mcf) and to be used for the thresholding of potentially unreliable low coherence areas.

For each burst overlap region an average unwrapped double-difference phase is calcualted. Only values with coherence values above the incidated coherence threshold are considered. Besides the average phase the program also calculates the phase standard deviation as an additional quality measure. Except for phase noise and phase caused by azimuth displacement a constant phase is expected. A significantly increased phase standard deviation is therefore a clear indication for the presence of unwrapping errors (which may occur considering that the high coherence areas may be spatially separated). Our experience shows that low phase standard deviation values (e.g. < 0.8 radian) are a quite reliable indication for a reliable average unwrapped double difference phase.

Besides the coherence and the phase standard deviation the program also consideres the fraction of the available non-zero pixels of the overlap region that was successully unwrapped. If this area is too small (e.g. smaller than a few percent) for a given overlap region then the average phase estimated for this overlap region is not further considered.

Such average double difference phases are estimated for each burst overlap region available. The averages for the individual overlap regions are then averaged using the a weighting factor calculated based on the validity fraction and the phase standard deviation.

This "global" phase offset between the earlier and later burst interferograms is then converted to an azimuth offset correction, considering the azimuth phase ramp present in the TOPS mode data.

The determined azimuth offset correction, is used to update the constant value of the azimuth offset polynomial (taken from the input offset parameter file and written after updating to the output offset parameter file.

Besides the updating of the azimuth offset polynomial the program provides a wealth of quality information on the estimation process, including coherence and phase statistics.

S1_coreg_overlap generates a large amount of intermediate files, including interferograms, coherence maps, coherence masks, unwrapped phases, etc. for every available burst overlap regions. Normally, these many intermediate files are directly deleted at the end of the program without looking at it in detail. In cases where the co-registration refinement result does not converge or where the quality number give indications of problems it is recommended not to deleted the intermediate files and check them carefully to be able to optimize the tresholds used. Deleting or not deleting the intermediate files is done by setting the cleaning flag to 1 (which is the default and which means the files are deleted) or to 0 to keep the intermediate files.

The factor for the scaling of the double difference phase offset determined to an azimuth offset in SLC pixels is around -0.016 azimuth SLC pixel per radian. This sensitivity also means that there can be a phase ambiguity problem for higher azimuth co-registration errors. A double difference interferometric phase of a full phase cycle (2PI) is reached for an azimuth co-registration error close to 0.1 azimuth SLC pixel.

OPTIONS

Using S1_coreg_overlap to refine the co-registration for a slave acquired with a long temporal separation from the master scene may result in generally reduced coherence. As a way to optimize the coherence it is possible to indicate in addition to the master burst SLC a second already co-registered SLC (possibly one close in time to the slave considered). The program uses then this second scene to calculate the double-difference interfergrams, while the master geometry is the one of the true master.

In cases where S1_coreg_overlap it may be necessary to start from already better co-registered burst SLCs. An initial quality of around 0.01 azimuth pixel should be achieved before running S1_coreg_overlap. Ways to achive this accuracy are matching techniques or to consider the overlap between the sub-swaths, as supported in the program S1_coreg_subswath_overlap.