Conceptually, the classification algorithm is reducing the floating-point, multi-channel images into a single-channel, low-bitdepth int.
An alternative to manually tweaking the parameters is to perform this computation using pure statistical methods.
This is similar to Maria Kavanaugh's seascapes.
bioDivMapR is an implementation of that approach which could be used.
A more advanced method would be to provide training points used to guide spreading of the spectral species classes.
Conceptually, the classification algorithm is reducing the floating-point, multi-channel images into a single-channel, low-bitdepth int.
An alternative to manually tweaking the parameters is to perform this computation using pure statistical methods. This is similar to Maria Kavanaugh's seascapes. bioDivMapR is an implementation of that approach which could be used.
A more advanced method would be to provide training points used to guide spreading of the spectral species classes.