When you have analysed your training sites and tested their separability, you are ready to run a supervised classification.
-
In the Maps tree, right-click Classification MetaLayer and click Run Classification.
-
In the Supervised Classification window, enable one of the following options in the Algorithm area
-
Parallelepiped: forces every pixel in the image to belong to one of the user-defined class types. If you choose this option and want to include Maximum Likelihood as a tie breaker, enable the With Maximum Likelihood as tie breaker check box.
-
Minimum Distance: forces every pixel in the image to belong to one of the user-defined class types.
-
Maximum Likelihood: allows a null-class parameter option. In some cases, you want to extract classes, but there are many more land cover classes represented in the imagery. Therefore, you want a proportion of pixels left unclassified, or null.
-
In the Classification Options area, enable any of the following check boxes:
-
Show Report: generates a report of the classification data.
- Save signatures
-
Create PCT: compares your classification with another classification.
-
Click OK.
The report should show a high overall training site accuracy. The information from each pixel in the training areas is compared to the information determined by the classifier algorithm. The overall accuracy represents the percentage of training-area pixels that were correctly classified. Your training areas are ideal examples of the classes.