CLASSIFYING AN IMAGE -- DENSITY SLICE

Remote sensing images are used to produce land use maps. Land use maps serve a wide range of purposes. In this exercise we will slice out water, wetlands, crops, and levees/bare soil.

The image used for this exercise is the Mississippi River. All seven bands of Landsat imagery are provided. They are all co-registered and have a resolution of 30 meters/pixel.

Landsat imagery represents reflectivity values of surface areas. Different types of ground features have different reflectivity. In order to classify raw satellite data ranges of values representing ground features are taken out of the image one at a time and assigned a specific value. This process is called density slicing. this is a primitive means of classifying images because we use only one band at a time.

In this exercise Band 4 (MISS4) will be used.

  1. IDENTIFYING FEATURES --Identify the following features on the satellite image.
  2. IDENTIFYING SPECTRAL VALUES OF GROUND FEATURES

    LOOK AT THE NUMBERS

    The task at hand now is to group data number values (reflected light) that are similar together. For example, water has low data number values because it absorbs lights. Using the density slice tool provided in NIH Image pixels with similar values can be grouped together.

    When this command is activated the group of pixels represented by the red bar, the pixels on the image, and the values on the histogram are all the same. The red band in the slicing bar is tricky to use. Using the LUT tool the band can be moved up and down and also increased or decreased. Move the bar around to get the feel of the tool. After the use of the tool becomes natural, adjust the bar and record the DN values for the following features. Identify the lower and the upper DN value.

    It is annoying, but the histogram has to be displayed each time the DN bar is moved. Compare the results with the ones listed below,

  3. SLICING -- SLICING -- SLICING

    In order to make a land cover map, the range of values selected above now have to be given one value per land type. For example, all values from 4-15 could be assigned a value of 20, all values of 16-30 could be assigned a value of 60, 31-47 would become 150, and 48-79 would be 200. This groups spectral values together and makes a land cover map that can be used to calculate areas from the imagery. This is an advantage not only of satellite imagery but of digital data.

    For centuries maps have been used to convey information. Not only should maps be as accurate as possible, they can be considered works of art. The colors, fonts, legends, and scale bars should not only be correct they should be as attractive as possible.

    The steps to make the land cover map are tedious and should be followed carefully.

  4. CREATE A NEW BLANK IMAGE -- it must be the same width and height of the satellite image. In this particular case the image width is 512 and its height is 512.
    1. File
    2. New
    3. width 512
    4. height 512
    5. map
    6. save as
    7. map
  5. DENSITY SLICE -- Move the red bar until the values that represent water are highlighted. In this case 4-15. Notice that the river and the other pixels representing water are red. Click on the value that will be assigned water. The tool looks like an eye dropper when you click in the bar. For example assign all the values of 4-15 a value of 20. Click on 20 and then click on the paint brush. Notice that the paint brush turns red. Now click the erase and choose what the background color will be. Choose white. The paintbrush should now be the shade you selected and the eraser should be white.
    1. Process
    2. Apply Lut
    3. OK

    The image now has the spectral values of 4-15 replaced with 20 and everything else is white.

    How could you determine the area of how much land is water?

  6. COPYING WATER TO THE EMPTY IMAGE MAP
    1. Double click on the dancing ants. This copies the entire image. Notice the ants dancing all around the image.
    2. Edit
    3. Copy Selection
    4. File
    5. Open "map"
    6. Edit
    7. Paste
    8. Click on anything else to get rid of the dancing ants.
    9. Save
    10. Go back to miss4.
      1. File
      2. Revert to saved.
      3. Options
      4. Invert
      5. OK
      6. Repeat the steps between the lines for wetlands. Assign wetlands a value of 60. When pasting the second image (wetlands)
        • Edit
        • Paste
        • Windows
        • Show Paste Control
        • Transfer Mode
        • OR
        • Click anywhere else to get ride of the dancing ants.
        • File
        • Save
      7. File
      8. Revert to saved.
      9. Options
      10. Invert
      11. OK
      12. Repeat the steps between the lines for urban. Assign cropss a value of 150. When pasting the third image (crops)
        • Edit
        • Paste
        • Windows
        • Show Paste Control
        • Transfer Mode
        • OR
        • Click anywhere else to get ride of the dancing ants.
        • File
        • Save
      13. File
      14. Revert to saved.
      15. Options
      16. Invert
      17. OK
      18. Repeat the steps between the lines for levees/bare soil. Assign vegetation a value of 200. When pasting the fourth image (levees and bare soil)
        • Edit
        • Paste
        • Windows
        • Show Paste Control
        • Transfer Mode
        • OR
        • Click anywhere else to get ride of the dancing ants.
        • File
        • Save
    11. CREATING A LEGEND, SCALE, and COMPASS BAR

      A map must have a legend, a scale bar, and a compass before it is complete. These map be drawn on the image directly or they can be made as new images and pasted on the finished map.

    12. MEASURING LAND TYPES

      In order to measure the area of different types of land cover area the scale of the image must be set.

      1. Using the measuring tool measure 1 pixel.
      2. Analyze
      3. Set Scale
      4. pixel
      5. known distance is 30
      6. unit is meters
      7. OK

      Calculate the area of the entire image. If this image is 512 width and 512 height and 30 meter resolution, the are in square meters would be 512 x 512 x 30 x 30 sq meters.) The next step is to calculate each type of land (water, wetlands, crop1, and levees/bare soil)

      1. Options
      2. Density slice
      3. Highlight Water
      4. Analyze
      5. Measure
      6. Show results.
      7. Repeat the above steps for wetlands, crops, and levees/bare soil.
      8. Record your answers.