GER 575/675 Exercise # 8

Name:_______________________________________

The final stage of classification (before accuracy assessment) is the labeling stage. You have already labeled your classes based on what you think the features on the ground are. You have used the signature plot, scatter plot and your knowledge of the ground cover in the local area to determine what the classes mean on the ground.

For this last exercise, you will label and recode your classes to generate a very general classification of the Logan image. To prepare for this, display your classified version of the Logan image and bring up the Attribute Editor. The last collumn of the Attribute Editor is Called Class_Names. For each class click on the appropriate space under Class_Names and enter the name that you gave to that class. When you are done, use the File/Save menu option to save your changes.

You now have thematic labels (the theme is ground cover) assigned to your spectral clusters. Your image has ceased to be a continuous image of brightness values and is now a thematic layer (or map) that that identifies the nature and location of ground cover. Your image is now a GIS layer complete with attributes.

However, you should have a number of your spectral clusters assigned to the same ground cover category. While this presents no significant problem, it would be nice if we had one category per ground cover type.

The process that redefines the coding scheme of your cover type map is called recoding. To recode a file means that the numbers that identify the types on the ground are re-numbered based on a logical scheme. In our case, if we want only three classes (water/shadow, vegetation, and bare surface) we have to identify which of our original 10 classes fit into these three categories and renumber the categories (recode).

Recoding in Imagine is relatively simple. We will use the three cover types listed above and recode our file into those three classes. Do the following:

  1. On a piece of paper make three collumns. In the first collumn list out the ten class numbers. In the second collumn write in the ground cover type that a specific cluster represents. In the third collumn write whether the class represents water/shadow, vegetation, or bare surface.
  2. Code water/shadow as class 1, vegetation as class 2, and bare surface as class 3.

    Under the Image Interpreter menu click on GIS Analysis. The menu bellow should appear:

    Figure 1.

    Click on the Recode option and the following dialog box will appear:

    Figure 2.

  3. In the Input File box click on your classified image and specify your Output File.
  4. Click on the Setup Recode button and the following dialog box should appear:

    Figure 3.

  5. In the first collumn called Value are your original 10 categories, the second collumn called New Value is the new category that you will recode your original values to (1,2, & 3). In the New Value collumn type in the appropriate slot the new value. Do this for all 10 classes. (0 should be recoded to 0 since it represents the background of your image and not a ground cover).
  6. When you have finished, click the OK button to return to the Recode dialog box.
  7. Click on the Ok button of the Recode dialog box.

    When the recode process completes, display your output file in the display window. The file should be completely black or it should have only three levels of gray.

  8. Bring up the attribute editor and add color to this file (make sure you save the changes).
  9. On a piece of paper list out your three classes, and the number of pixels associated with each class (Histogram). Each pixel is 30x30 meters in size. How many hectares are contained in each class?