Activities‎ > ‎Investigations‎ > ‎Remote Sensing‎ > ‎

Using AnalyzingDigitalImages to identify features in a Landsat JPEG

Jump to [materials] [what to do]

Rita Freuder From a Teachers Workshop, June 2006, Portland, Maine

Image to use for this experiment
The students were using AnalyzingDigitalImages on the JPEG format of a Landsat 3-band false color image to look at colors within limits defined by a user box. We all made some discoveries. If the box is drawn with the most uniformity of color in mind, it may tell you about landuse in the image. For example, in a false color image, (Landsat bands 4-3-2) a bright light blue is usually concrete, or buildings or other infrastructure. It turns out that the residential areas and highways are better defined by the blue-grey areas.


Materials: (1 & 2)

    1- Right click on this image to download.

    2- Download the AnalyzingDigitalImages software as part of the DEW software bundle from the software download page.

    3- Double click the AnalyzingDigitalImages application. Use the 'Open a Picture' button in the upper left of the screen to load the jpeg file from step 1. It is Portland, Maine - ETM2000.jpg, from an Aug 26 2000 Landsat image.

    4- You can define a scale or not. Say none for now. The unit of pixels will work since this jpeg was made from a 512 by 512 pixel data input. So the percent of the image that matches the user-defined box of 3-color bands could be computed in square meters if desired, since one Landsat TM or ETM+ pixel is 28.5 m on a side.

    5- Now that you have the image on the screen, try this example:

What to Do...

  1. Go to the 'Spatial Analysis' tab panel and select the 'Rectangle Tool' from the drop-down menu.
    2.
  2. Draw a box within the darkest red color (probably conifers)
    3.
    Screenshot of AnalyzingDigitalImages to draw a rectangle on the image

     
    To draw a rectangle in AnalyzingDigitalImages, click on the mouse for the upperleft corner of the area you want to highlight and drag the cursor. When you release the mouse button, the red dot is the lower right of your box.
  3. In the 'Mask Colors' tab panel, you can choose a range of colors to display. The default is RGB. The intensities of this color should appear in the color bars below the heading on the left.
    4.
  4. Draw a box over the area you are going to measure. Near the bottom left side of the screen choose: 'Apply Mask-TEST 1' or 'Show Mask.'
    5.
    Applying a mask on the image in AnalyzingDigitalImages. Now black represents the pixels that match the RBG values of the mask

    A black and white image appears where black represents the pixels that match the range of RGB of those in the box. The white is the unmatched pixels. The number of masked pixels can be viewed in the 'Spatial Analysis' tab panel. To view the percentage of masked pixels, select 'Save Measurement' from the Measurement menu and follow the instructions to save your data. The percentage of masked pixels will appear in a window like the one below.

    Viewing the percent of pixels that match the specified mask

  5. Choose Show Original Image to help clarify what the black area is representing. Alternatively, Mac users can press CTRL + H.
    6.
    Using Analyzing Digital Images to contrast the original image with the masked image. Specific pixel values (colors) are turned into black and everything else goes to white.

  6. To draw another box, first choose Display Tools: Red-green-blue, then draw a new box (steps B->E)
    7.

    Now decide on what other color/landuse you may like to see. I suggest deep ocean, light pink areas, the cove area, bright blue in urban area, and grey-blue in urban area, and medium red. I noticed that the sum of all the "color types" did not add up to 100%.What landuse do you think these boxes represent -- e.g. lakes and rivers, etc ?