search

iprapas / landslide-sar-unet

Repository for the paper "Deep Learning for Rapid Landslide Detection using Synthetic Aperture Radar (SAR) Datacubes"
MIT License
32 stars 3 forks source link

Details about the variables #4

Open cOsprey opened 4 months ago

cOsprey commented 4 months ago

The datacube has topophase_flat_imag, topophase_flat_real, topophase_cor_0, topophase_cor_1 can you explain or provide more details about them? What is meant by flat and cor_0 and cor_1?

iprapas commented 3 months ago

If i remember correctly, these are the insar phase calculations. Flat imag(inary), real should be the difference in complex form, and cor should have the coherence. Not sure about the difference between cor_0 and cor_1. @weiji14 could you enlighten us maybe?

weiji14 commented 3 months ago

Yes, these are the interferometric products produced from ISCE2 (we had documented some details at https://gitlab.com/frontierdevelopmentlab/2022-us-sarchangedetection/deepslide/-/blob/main/notebooks/README.md#13-insar-processing-using-isce2-data_isce_processing). The original output from ISCE2 is a file called topophase.cor.tif, and I'm having trouble remembering what cor_0 and cor_1 means too (the ISCE manual isn't very helpful here) 😅. A quick search yielded this page - https://adenaheem.com/standard-procedure-for-generating-differential-dems-from-insar-using-isce-strimapapp/ which says:

topophase.flat: flattened (geometrical phase removed) and multi-looked interferogram.(one band complex64 data). topophase.cor: coherence and magnitude for the flattened multi-looked interferogram. (two bands float32 data).

Hopefully this is a meaningful explanation?

cOsprey commented 3 months ago

@iprapas and @weiji14, thank you. I am confused about cor_0 and cor_1, as generally the coherence should have value between 0 and 1, which is correct for cor_1 but cor_0 has values more than 1, seems something else.