I am performing an image classification(Random Forest) for obtaining rice and non rice areas from Sentinel-1 data. As a result am obtaining same classification accuracies for both polarizations. I wanted to know whether is it right to get same answers irrespective of polarizations used?

  • Just a comment: A random forest classifier is based on the principle of random permutation of both the input features (your rasters) and the trainings samples. If you only use one band (VV or VH if I understood you correctly), the randomization of the input features is useless because every tree you compute is based on the same raster values, regardless of the number of trees. This takes one of the largest advantages of your classifier. Consequently, you have to make sure that at least the number of training inputs is large enough for shuffling so at least there the randomization is effective – AndyB Jul 12 '19 at 4:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.