Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It only takes a minute to sign up.
Sign up to join this community
I am training a convolutional neural network to classify landcover and I would like to ingest both Landsat 8 and Sentinel-1 into my classifier.
Currently it is working reasonably well for Landsat 8 but I want to get Sentinel-1 data, ideally at the same resolution and pixel locations. Is there any library that already does this or can you think of a good approach ? I'm using rasterio in python for most of my raster processing.
I'm downloading my Sentinel-1 data from Sentinel-Hub.
Take a look at rasterio's virtual warping. You could e.g. take the higher resolution Sentinel-1 image as a template raster from which you get the "vrt_options" dictionary with a specified crs, transform, pixel width/height etc. Then use that to read in both the Sentinel-1 and Landsat images, rasterio warps them on the fly to the specified options (be aware that this forces resampling of the lower-res/different grid Landsat images).
If you work with the snappy modules (libraries of SNAP) you can use the collocate module to resample Sentinel-1 onto the resolution of Landsat-8. But the S1 data is delivered as GRD product which means that it has to be geocoded first. In case of 30 m are sufficient, multi-lookig to 30 meters is also an option. Depends on how far pre-processed the Sentinel-1 products you use are.
An example for its use is given here: https://eodag.readthedocs.io/en/latest/tutorials/tuto_burnt_areas_snappy.html
