IDENTIFYING DEFORESTED AREAS IN AMAZONIA USING DIGITAL IMAGE PROCESSING

ABSTRACT

This work presents the methodological sequence for identifying and mapping the deforested areas, in Amazonia, using digital image processing of Landsat TM data. The spectral linear mixing model was used to generate vegetation, soil, and shade fraction images derived from the original bands of TM acquired in 1997 and 1998. An example of this approach is presented in area of Rondônia State, characteristic of the human occupation in a fishbone pattern. The shade fraction image of 1997 is utilized for mapping the extension of the deforestation (including areas of deforestation, pasture, and forest regrowth). This mapping is performed using the image segmentation and non-supervised classification algorithms. Hence, these deforested areas are overlaid on the soil fraction image of 1998, allowing to mapping the increment of deforestation using the identical classification procedure. The result of these both classifications, in digital format, can be included in an Amazon database. Currently, this method is being used, in an operational way, for 47 TM images that cover the critical area (PROARCO - IBAMA) in terms of deforestation in the Amazon region.

Key words:
Amazonia; Remote Sensing; Deforestation; Tropical Rainforest; Image Digital Processing