|Exercise 2: Land cover classification|
Now that you have become familiar with the Córdoba area, we are going to create a land cover map. This will be done by supervised classification, meaning that the information to map the image for different land cover classes will come from an external source. An algorithm in the software will compare the values in each pixel with the information provided, and will calculate to which class the pixel is most related.
We will do this twice: once for the 2009 image, and then for the 1992 image. In the previous exercise we identified three main land cover classes which will be used in this exercise as well. Such an approach will allow us to quantify each land cover class and assess the surface of the urbanised area, both in 1992 and 2009.
Land cover classification
Open/Multiple Files as Single Dataset: 2009_cordoba_B1 to 2009_cordoba_B7
For LEOWorks3: Open the 2009_cordoba_B1.tif image
In the Image Preview window click OK.
Now repeat the process for the other 2009 bands.
To perform a supervised classification, we need predefined information. This information was provided to you in the form of the files class Urban.shp, Bare soil.shp and Vegetation.shp (LEOWorks4) trainingfields2009 (LEOWorks3) We are first going to investigate these training fields. In order to do this, we need an RGB image.
Display band 4 in red, band 5 in green, and band 3 in blue, in order to create a false colour composite, just like you did in the previous exercise.
Go to Tools - Classification – Supervised - and create 3 classes: Urban, Bare soil, and Vegetation. Import the training fields TF from shapfile (see the LEOWorks Tutorial for more detailed instructions).
For LEOWorks3: Select Multivariate Analysis- Supervised Classification-Select Trainingfields-Open Trainingfields: trainingfields2009.