In [ ]:
%matplotlib inline
Here are the RadarSat-2 quadpol coherency matrix image directories as created from the Sentinel-1 Toolbox:
In [ ]:
ls /home/imagery
To combine the matrix bands into a single GeoTiff image, we run the python script ingestrs2quad.py:
In [ ]:
run /home/ingestrs2quad /home/imagery/RS2_OK82571_PK721079_DK650144_FQ17W_20160403_230258_HH_VV_HV_VH_SLC/
In [ ]:
run /home/ingestrs2quad /home/imagery/RS2_OK82571_PK721080_DK650145_FQ17W_20160427_230257_HH_VV_HV_VH_SLC/
In [ ]:
run /home/ingestrs2quad /home/imagery/RS2_OK82571_PK721081_DK650146_FQ17W_20160614_230256_HH_VV_HV_VH_SLC/
Here is an RGB display of the three diagonal matrix elements of the above image (bands 1,6 and 9):
In [ ]:
run /home/dispms -f /home/imagery/RS2_OK82571_PK721081_DK650146_FQ17W_20160614_230256_HH_VV_HV_VH_SLC/polSAR.tif \
-p [1,6,9]
To estimate the equivalent number of looks, run the python script enlml.py:
In [ ]:
run /home/enlml /home/imagery/RS2_OK82571_PK721081_DK650146_FQ17W_20160614_230256_HH_VV_HV_VH_SLC/polSAR.tif
So the ENL would appear to be about 5.
To run the change sequential change detection on the three images, run the bash script sar_seq_rs2quad.sh. It gathers the three images together and calls the python script sar_seq.py which does the change detection. By choosing a spatial subset (in this case 400x400), the images are clipped and co-registered to the first image. This might be unnecessary if the images are well registered anyway.
If you have a multicore processor you can eneable parallel computation by openeing a terminal window in the container (new terminal) and running
ipcluster start -n 4
In [ ]:
!/home/sar_seq_rs2quad.sh 20160403 20160427 20160614 [50,50,400,400] 5 0.01
Here is the change map for the most recent changes:
In [ ]:
run /home/dispms \
-f /home/imagery/RS2_OK82571_PK721079_DK650144_FQ17W_20160403_230258_HH_VV_HV_VH_SLC/sarseq(20160403-1-20160614)_cmap.tif -c
In [ ]: