To avoid distortions in SAR pixel values due to geocoding procedures, it is sometimes recommended to extract features in slant range geometry. However, finding ground-data points/plots in slant range scenes may be difficult for natural ecosystems…, and this task can’t be automated when analyzing multitemporal data. Thus, I’d like to convert known lat/long or UTM coordinates to the slant range geometry of SAR images.
I found a mention of cartographic coordinates to slant range geometry conversion in the manual of a propietary SAR processing software. I had also read about reverse geolocation algorithms in RADARSAT-2 technical notes, which suggests using orbit data or rational functions.
I think this function is not available for SNAP S1TBX. Do you know some script, plugin or open source software to do this? Or may be this type of algorithms are not available because the results would be incorrect or inaccurate?
SNAP can actually import vector files and masks (e.g. land/sea mask) on non-Terrain Corrected images, so somewhere in the code there must be a conversion from geographic to image coordinates (slant range or ground range). I don’t know which method is used to do this conversion. However, in my experience (importing accurately geocoded shapefiles on non-TCed Sentinel-1 images), this conversion can be very inaccurate (off by several hundreds of meters).
I am not familiar with rational functions, but geolocation using orbit data will be accurate, assuming of course that the orbit data is accurate.
SNAP uses the orbit data in the Range-Doppler Terrain Correction function, but that’s forward geolocation. The orbit interpolation step is shared between forward and reverse geolocation, in case you want to use SNAP’s code as a starting point.
I don’t know whether open source software to do what you need exists. It would be nice for SNAP to have this functionality. Currently it doesn’t have it AFAIK.
Hi, thanks for your response!
It would be nice to add the reverse geolocation functionality in SNAP!
How do you import vector files and masks on non-terrain corrected images? A displacement of several hundreds of meters is very large for me too…, but I may explore that conversion.
Thanks! I could import a vector on the slant range image
I attach some images (white lines correspond to the vector file): I had to flip the ALOS/PALSAR-1 scene (using the GraphBuilder). However, the vector was also flipped. I tried to replace the metadata of the flipped scene before re-importing the vector file. Since this not worked, I “manually” replaced the metadata using the band math trick mentioned here. Lastly, the vector was correctly oriented.
As @css said, there is a displacement of ca. 300 m (the vector should fit on the riverlines).
May be geolocation using orbit data can improve this result. For the moment, this procedure helps me to draw polygons (and find known targets) on the slant range geometry. Thanks!
To update this post: importing vector files on slant range geometry works well on RADARSAT-2 imagery (SNAP 7.0, Linux version), and there is no need to replace the metadata.
