I am encountering some problems in terrain correction through snappy. I perform an ocean object detection processing chain through snappy. It seems that adaptive thresolding and object discrimination work fine. I have an output of two bands, a sigma0_VV band and a sigma0_VV_bit_msk band.
The first question is: is it true that the help section of ship detection gui is old? I cannot see an avaialble layer with object detection results
The second problem is that, when I perform terrain correction, I obtain the following error:
SEVERE: org.esa.snap.core.datamodel.Product: raster width 845 not equal to 1240
where 1240 is the dimension of the raster image before terrain correction, and 845 is the raster dimension of geocoded image after terrain correction.
Moreover, I can see the ShipDetection mask as result, but the image band has no value. The option noDataValueAtSea is set to ‘false’
Could it be there is an error in using the option noDataValueAtSea? I say this because the image shows some pixels near the coastline, where tha land has been previously masked. So it seems that the sea is always masked during terrain correction execution
Yes, I have tried the same processing with the GUI, and through a bash gpt command executing an xml graph too.
Using the GUI I have no problem. Both Ocean Object Detection and Terrain Correction steps are performed with no errors.
Doin some other tests, I noticed the following strange behaviours:
Using gpt bash command and xml graph
The terrain correction step produces the raster mask and the vector with object detection results only if, after Object Discrimination step, I write Object discrimination result on disk, and then run Terrain Correction starting from the written file. If I process the graph going directly from Object Discrimination to Terrain corretion no masks or vector are produced (this was working fine with SNAP 4.0; now I am using SNAP 5.0)
The terrain correction step produces the raster mask and the vector with object detection results only if, after Object Discrimination step, I write Object Discrimination product on disk, and then run Terrain Correction by using Object Discrimination data still in memory, without accessing to the just now written file. I.e., if I perform the processing going directly from Object Discrimination to Terrain corretion, or using the Object Detection product written on disk, no masks or vector are produced
Regarding the fact that the image band has no data value (in all the processing performed with sanppy), may it be there are some issues with SRTM use? I noticed that when the process starts downloading the SRTM DEM, a wrong tile is downloaded (42_06 instead of 41_06). I tried to set the saveDEM option in Terrain Corretion to “true”, and elevation band too has no data value
I hope I was clear enough. If you need some more explanation I am at your disposal