The content of a hyperspectral image can be classified and also quantified using PLS-DA. The following scheme is possible for doing this:

1. Create a PCA model of the image data.

2. Create an RGB image and a Scatter 2D density plot from the PCA scores, T.

3. Make selections in the Scatter 2D density plot that corresponds to a class (a set of pixels with similar properties) in the RGB image. The first selection in the used example file looks like this:

4. Right-click on the selection in the Scatter 2D plot and choose "Set selected to: New category". Choose to create a class from selection in the dialog that opens.

5. Make the next selection in the Scatter 2D density plot that corresponds to the next class.

6. Right-click on the Scatter 2D plot and set the selected pixels to a new class.

7. Right-click once more on the Scatter 2D plot and choose "Selections" from the plot menu. Mark the two classes shown in the dialog and press the "Select" button found at bottom left. Now, press the "Invert" button, which means that all pixels not belonging to the two created classes will be selected. Right-Click once more and set these pixels to a new third class.

8. Three variables will now be visible at the bottom of the DataSet table when viewing the variables. Set these to Y.

9. Create a PLS-DA model from the DataSet in the Data Tree.

10. In order to simulate an unknown sample image, set the a certain part of the image to a test set by first selecting and then right-clicking on the DataSet table while choosing "test". It could also be possible that this part of the image contains an "unknown" area, for example if the image is a composite image, i.e. comprised of several images stitched into a single image.

11. Create a prediction table by dragging the PLS model onto the table area. The prediction tables gives a classification as well as a quantification of the image test set content.