Characterization of Cerrado Vegetation Using Automatically Classified Landsat MSS Data

摘要

The characterization of physiognomic unities of cerrado vegetation through automatic IMAGE 100 analysis of LANDSAT MSS data was studied. The separability of the unities analyzed was verified (cerrado, campo cerrado, campo sujo) from respective spectral signatures obtained using a MAXVER and a JM distances algorithm. A correlation analysis of correctly classified values was performed. The results obtained include the following: (1) the use of four channels together presented the best class separation, independent of season; (2) in considering the use of isolated channels, channel 5 provided the largest percent correct classification; and (3) the highest separability was observed between cerrado and campo sujo forms. It is concluded that the separability of physiognomic unities can contribute to studies for the rational use of cerrado areas.