Training data classification

Hi all,
I want to classify my geometry training data, so I am trying out the supervised classification with random forest. But the result show that all of the area have been classified. I mean, could I just classify my geometry training data not all of the area?


Hi @ninis,
Can you show the processing graph?

Hi @adraghici thanks for your respon,
Here my processing pic. I just want to classify the red mask on that pic.

But the result show all of the area also classified with this RF classification, as you can see on the 2nd pic.

How can I classify only on masking pixels?

You can mask out the rest of the image using the geometries, they are stored as masks automatically using the MaskManager. Still the whole image will be classify, but you use the mask later to hide the non-valid pixels.

oh i got it, thank you. I have new question again. Could I clip the area by poligon/vector in SNAP ?

Hi, this has been asked several times in the forum already.
With a workaround it is possible.
In a nutshell:

  1. Use the bounding box of the polygon to create a subset
  2. Use the LandSeaMask operator to mask pixels outside the polygon

Please have a look at the following posts and the linked threads.

thank you mario and ana.