Cordoba 2009 with training fields

Page12

For LEOWorks3: Use the Inspect TF Statistics and select from the table which TF you want to inspect on the image.

Now you can see which pixels have been selected as training fields for the supervised classification. All the small marked areas on the image present pure land cover classes, of either urban land, vegetation or bare soil. These training fields have been created by selecting pixels that contain only one land cover. After extracting the information contained in every band of these pixels, a (statistical) signature of each type of land cover and sample is created.

The algorithm needed to execute the classification is called ‘maximum likelihood’. This process assigns each pixel of the entire image to the land cover class it most likely belongs to, based on the a priori information and statistical measures.

Start the process of Classification. Select all the bands of 2009_cordoba images. Set the threshold at 0, so that all pixels are classified.

 
 

	Example of a classified image

Add the legend and eventually change the colours of the land cover classes. Study the Report (No. and area of pixels in each class). You may produce the following output:
 
 

Save this classified image as Cordoba_2009_class, but keep the RGB image on your screen, as we will need it for the next exercise.

Page12