Directions: Supervised Classification
 

Home page

Introduction

Using Remote Sensing

  1. Obtaining
  2. Linking
  3. Subsetting
  4. Compositing
  5. False coloring
  6. Ratioing
  7. Supervised classification

Examples

Pre-post test

Directions for making a supervised classification of a LANDSAT image

This process groups similar land cover areas so that a more quantified analysis is possible. Reference on web: http://www.geo.utep.edu/pub/seeley/Classification_tutorial_multispec.html

Background requirements: Downloaded and familiarity with the free MultiSpec© software and tutorial (See http://cobweb.ecn.purdue.edu/~biehl/MultiSpec/Index.html) and a basic understanding of satellite images is helpful (i.e. bands, false color imaging, resolution).

Refer to the direction sheets titled: Directions For Obtaining Ohio LANDSAT Images, Directions for linking LANDSAT single band images, Directions for Subsetting LANDSAT images.

  1. Open a downloaded, linked (subssetted, and composite if desired) LANDSAT image.
  2. Click File. Select Project. Select New Project. Click OK on next window.
  3. Highlight box “new” in Class box if not highlighted.
  4. Draw a small rectangle around a homologous area.
  5. Select “add to list”.
  6. Enter a new class name that represents your training area, such as “water”, click OK .If you want to differentiate this class you may create fields within the class. I.e. “deep water, shallow water”.
  7. Select “New” in the Class Field Class Field box.
  8. Repeat steps 4-7 as needed.
  9. Select Processor. Choose Classify.
  10. Select “Window Image Overlay.” Click OK on next window.
  11. Save your classification with a convenient appropriate name such as beginning with sc for supervised classification.
  12. To determine area of each classification, click on “text output” (window behind your active image).
  13. Validity: Check the validity of your classification through ground truthing or visual comparison of similar fields before and after classification.