Using shade fraction image segmentation to evaluate deforestation in Landsat Thematic Mapper images of the Amazon Region

Abstract

Image segmentation based on the shade fraction of a Landsat TM image was effective in measuring the areal extent of Amazonian deforestation. The shade fraction image derived from spectral mixture models was related to the forest canopy structure. Dense tropical forests have a medium proportion of shade within their canopy while deforested areas (bare soil, pasture, and/or regrowth) have a comparatively smallproportion. Comparison of image segmentation results with conventional techniques showed visual agreement. Even though additional tests are necessary to validate this approach for large areas, the technical soundness of the approach has been demonstrated.