USE OF THE SPECTRAL CLASSIFIER CORRELATION MAPPER IN TM-LANDSAT IMAGES

Abstract

In this work a methodology is proposed for the analysis of spectral mixture from TM-Landsat data. The selected area in the image embraced preserved and agricultural areas in Brasília, DF. The adopted procedure for endmembers’s (pure pixels) detection was the same used for the hyperespectrals data treatment, with the specific adaptations. Initially was made a conversion of the Digital Numbers for reflectance values and after, the identification of the endmembers of the selected area. The Digital Numbers conversion in reflectance values allowed the comparison of the spectra obtained with the spectral libraries, which could help in the identification. For that area, the identified pure members were: green vegetation, non-photosynthetically active vegetation, soil and burned. It was used the spectral classifier SCM (Spectral Correlation Mapper) to show the pure members disposition in the image. The use of the endmembers detection procedure before the classification revealed quite interesting because it helped in the samples determination for the classification. Using the analysis of spectral mixture the classification becomes more trustworthy to the pixel conception that can contain more than one class.