Additional interferogram network options for Sentinel Stack processor

[falkamelung]

stackSentinel.py currently offers for network generation only the --num_connections option. For example, --num_connections 4 will create for each date 4 nearest-neighbor interferograms.

It would be good to have the ability to skip short-term interferograms and to add several interferograms with at least 1 year temporal baseline.

I'd suggest to change the --num_connections option as follows:

--num_connections #_of_connections  time_lag
--num_connections #_of_connections1  time_lag1   #_of_connections2  time_lag2

With this

--num_connections  4 40

would create 4 connections with 40 days temporal baseline and more (0-40, 0-52, 0- 64, 0-76_ day interferograms. The option

--num_connections  4 0  2 180 1 365

would create 4 short-temporal baseline interferograms, 2 180-days or longer interferograms, and 1 365-days or longer interferograms Obviously,

--num_connections  4 0

would be the same as --num_connections 4, i.e. we would have full backward compatibility.

We should should add options such as --delaunay with a scaling parameters.

Let me know your thoughts and about any other options that need to be implemented.

[hfattahi]

Great idea. In general we could have a network selection function which we could use here. Looks like networkx has some nice functionalities. Also we have some select_pairs functions in mintpy that we could reuse here and build on.

[falkamelung]

Thank you for the suggestion. NetworkX looks great, but the MintPy function is probably easier to implement. Assuming that we don't want a dependency of ISCE on MintPy we probably will copy the network.py from MintPy to ISCE and commit there. Let me know if you seen anything that could be of concern.

[mjacqu]

I would love to see this functionality added, as it has been one of the main issues I've had with the stack processor. I implemented something along these lines for automating the pair creation with topsApp, where I could define a maximum time difference allowed for the formation of date pairs: https://github.com/mjacqu/ISCE/blob/master/tops.py make_pairs function (example here: https://github.com/mjacqu/ISCE/blob/master/topsrun.py) I think ideally one could define both the minimum and maximum temporal baselines, and all pairs matching that constraint would be run.

[ThreeIcug]

In the GMTSAR， they have an script to select pairs according to the given threshold in time and baseline(select_pairs.sh). I think maybe it will help us. But in the first we need to calculate the baseline.