supervised classification

[Greg Gaston]

I have been working with i.cluster and i.maxlik and GVI data to create unsupervised classifications of the former Soviet Union.  I would like to experiment with a supervised classification. How do I create training site signatures to use as "seeds" in i.cluster? 



Thanks

greg gaston 
gregg at russia.cor2.epa.gov


  I am also still in the dark about the units of the "class separation" parameter in i.cluster... any thoughts?