You can initialize a new classification session with or without data files open.
In the Session Configuration window, type a name in the Description box.
It is not necessary to identify features in the image in unsupervised classifications. Unsupervised classifiers do not use training sites; they divide the image pixels into natural groupings of statistically similar gray-level pixel values.
If you want to work with more than one algorithm in the same session, use the Open command on the Select Algorithm window. The Select Algorithm window remains open until you click Close.
Adjust the Red, Green, and Blue color values to the combination you want by clicking the appropriate column for each channel.
A project can contain several classifications, each one using different sets of input channels.
In the Input Channels column, select channels you need for your project.
Select an output channel to store your classification by clicking in the Output Channel column for the channel you want to use.
You can select an empty channel or you can over write an existing channel, such as a channel containing imagery of no consequence.
If there are no desirable or available output channels, click Add Layer and enter values for channels in the Channels to add boxes for the appropriate channel types. Click Add.
Click OK.
The Unsupervised Classification window opens. Focus adds a classification metalayer to the Maps tree and the RGB reference image you specified opens in the view pane.