Problem in classification by adding images by linear stretching

Microwave Toolbox
5

  1. The random forest classifier actually doesn’t care about different units and value ranges of input data.
    What do you mean? Do you mean, we do not need to do ‘signal stretching’ for random forest?

  2. But if you really want to transform them, you can convert the whole product to uint8 (resulting in values between 0-255) and divide the product by 255.
    How can I convert to unit 8, is there any option in bandmath for it?

  3. What do you mean by “the classification does not work good”?
    It mixes between one class and NAN values like here:
    Classification mixing problem after using PCA
    I do not have this problem in original bands. This is strange.

  4. After you use “convert band” you have to save your product again.
    I know I saved it but still does not work.