Maximum likelihood classification using only certain bands


I would like to do a supervised classification of my product using only certain bands as input and using maximum likelihood to assign pixels. Is this currently possible ? I couldn’t figure out how to do this in the app.

Actually I don’t really understand how the SNAP classifier identified as Maximum Likelihood (in Raster -> Classification) operates and what it outputs. Could someone shed some light on this for me ?

Thank you !


First, produce some features. These could be the sigma0 backscatter, the GLCM textures, coherence, polarimetric decompositions, etc.
You can select only certain bands to use by selecting them in the “Feature bands”.

To train your classifier on vectors, you can either draw your own and label them or import a shape file as use the names or attributes as labels.

The results are the predicted classes based on the feature bands and the training vectors. The colours of the classes are randomly selected. If you are classifying land cover for example, you will need to import your own colour palette or select the colours for each class.

Right ok, I must have been pretty unfocused there because I missed the scrollbar in the MLC window and just didn’t see the Feature Bands section.

Thanks a lot for your patience !