Absolute Geolocation Error for CSLC-S1

[vbrancat]

This PR aims at evaluating the Absolute Geolocation Error for OPERA CSLC-S1. The script ale.py requires as input:

1. A CSLC-S1 product
2. A CSV file containing the surveyed Corner Reflector locations (CR)

For each CR location contained in the CSLC-S1 product, the code:

1. Identifies the peak corresponding to the CR and computes the image coordinates
2. Computes ALE by differencing the computed CR location with the surveyed CR location

Results can be saved in a CSV file on disk or displayed on screen with a plot

[vbrancat]

@seongsujeong and @hfattahi the latest commits have: 1) Syncronized peak and geolocation error detection with Jeong's algorithm available here: https://github.com/seongsujeong/compass_util/blob/3aa086cdce903ce3853557592100563b69ceb05b/compass_util.py#L122. I have checked that the two codes produce the same geolocation error and they do. I have verified this on multiple bursts at multiple dates. Happy to point you to the data location 2) Expose more options in terms of polarization, margin, oversampling factor etc 3) Try to generalize the code to read NISAR GSLCs. @hfattahi it would be great if you can point me to a NISAR GSLC so I can test the NISAR part of the code a bit more carefully

Let me know your feedbacks

[vbrancat]

Thank you, @hfattahi! I addressed your comments. It would be great if I can have a toy GSLC so I can test my changes. the test on OPERA CSLC is pending and needs to be performed for the last set of changes

[seongsujeong]

Tested the code in this PR by taking a look at CR05, and comparing the result from track_cr_geogrid.py in here

python ale.py \ -p t064_135523_iw2_20221016.h5 \ -c 2022-10-16_0151_Rosamond-corner-reflectors.csv

ALE_X: 1.456947, ALE_Y: 0.474894

track_cr_geogrid.py \ -p -118.070803 34.80549368 661.2381 \ t064_135523_iw2_20221016.h5

error=[1.456033, 0.474903]

I think the result from this PR is consistent to mine. Nice work.

[vbrancat]

@seongsujeong not sure why I am not able to comment on your posted comment, do you think that a fixed margin would do the trick? If that is the case, you much would you recommend?

[seongsujeong]

@seongsujeong not sure why I am not able to comment on your posted comment, do you think that a fixed margin would do the trick? If that is the case, you much would you recommend?

I think fixed margin would be okay as long as we set up enough margins around the peak. In my pessimistic estimate I would say about 6m in geolocation, about 3 pixels in peak radius, (so, 3 sqrt(5 10) = 21.2), and 1.5 of safe factor. That would be about 33m, so I would suggest 35m of margins.

[vbrancat]

@seongsujeong all your comments have been addressed :D