Ah, the "Simple" method

If you'd like to see what this process is like open the example algorithm found in
ermapper/algorithm/Example_Classification_Compute/ directory called
Compute_vegetation_over_RGB_541.alg
and look at how the algorithm is created.

So you've got a ratio that seems to highlight what you'd like it to. How do you turn that into a classification layer?

  1. Find out what the values are by choosing Cell Values.. from the View Menu.
  2. Select the pointer from the main toolbar click on the image. In particular you want to see what the various bands, ratios, or PC's are giving you at these distinguishable points.
  3. Add a new "classification" layer to the list.
  4. Choose the formula editor and enter an equation that codifies what your image shows you

    5. if input1>1.25 then 1 else null
    or
    if input1/input2>1.25 and input2/input3>50 then 1 else null
    etc
    (null lets you see through the layers, "else 0" replace all empty cells with white)
  5. Fromt the color bar on the process stream, choose
  6. Change the inputs to match the bands, ratios or PC's that you're using and click GO. If you're using PC's you'll have to save the PC algorithm in a different format using Save As a Virtual Dataset before you can use it in the classification layer.
  7. The "classification" layer should show you all areas that satisfy your conditions. Change them as required.
This works very well for IR bands and band ratios in Virginia. For example, here is a chunk of an image I created (look at small_Va_RGB_7:5_4:3_3:1.alg in my home directory) that uses three ratios 7/5 4/3 and 3/1 for RGB respectively

Why does the town look red. Either the 7:5 ratio is very high or the other two are low. If we look at the cell values profile under the edit menu and click on a part of Lexington we get this.

notice how the 7:5 ratio is not very high (but 6 is very high, lots of heat!). But the 4/3 and 3/1 ratios are both less than 1. So red wins. Compare that to the green areas, where 4/3 is very high.

A simple class for the urban areas might read
if input1/input2 > 1 and input3/input4 < 1 then 1 else null
given
input1 = band 7
input2 = band 5
input 3 = band 4
input 4 = band 3
might show you where the urbanized area is.