CD workflow in phases

I want to do Change Detection on a library of bursts (that all have the same footprint), which would be ~30 times smaller than the corresponding library of full S1 files.

For full S1 files, my normal S1 CD workflow is this:

It seems like it should be possible to separate this into two steps - one that splits out the bursts of interest from the large files, thus creating the burst library:

Followed by one that operates on the bursts in the library to do the CD:

However, when I try this, the back-geocoded secondary image is always filled with NaNs. Others on this forum have reported similar problems, but I can’t track down why this might be the case - it happens with various kinds and resolutions of DEMs, all of which are being successfully downloaded.

So is there some basic reason why back geocoding won’t work with individual bursts?


Think I found the fix. Doing it in two steps is fine, but TOPSAR Deburst should move from Step 1 to Step 2:

Step 1, wherein we build the library of bursts:

Step 2, wherein we do cool CD stuff:

It might still be faster if you do terrain correction separately.