Tuesday, April 13, 2010

Landsat Satellite Images Change Detection Methods

Minnesota's Department of Natural Resources (DNR)'s website depicts its use of Landsat Thematic Mapper (TM) images to map Minnesota land cover. Landsat TM images are digital and differ from traditional digital images in their ability to express additional measures of brightness beyond simple RGB brightness. Specifically, Landsat records an additional four sets of brightness from the near, middle, and thermal infrared portions of the electromagnetic spectrum. Landsat images are useful in showing changes between two images taken of the same area but years apart. Image differencing, or subtracting the original image from the new image to determine changes, is a simple technique in remote sensing.

Preparatory Steps

Both images should show the same season and must be accurately registered (matched up) to the ground and to each other. Remove areas with clouds and distinguish between forest and nonforest areas. The images must be radiometrically calibrated to minimize effects of instrument variations and atmospheric haze.

Analysis

Values in the difference image that are red or orange show loss of vegetation, while greens depict growth of vegetation. The example uses a simple bell curve in which minor changes in vegetation account for 80% of the total changes and significant changes occur on either side of that distribution.

Strengths

  • Relatively easy to perform
  • Objective in nature

Weaknesses

  • Chance for error if preparatory steps not followed
  • Does not account for other factors of vegetation change between the images
  • Unlike satellite image analysis, it is not possible to compare analysis units and determine which has shown the most vegetation loss in this case, since the statistical thresholds display roughly the same amount of change and non-change regardless of what has actually happened in the area.

Source: http://www.ra.dnr.state.mn.us/changeview/change_tech.html

No comments:

Post a Comment