COMPARISON OF CLASSIFIER SUPPORT VECTOR MACHINE AND SPECTRAL ANGLE MAPPER APPLIED ON DIFFERENTIATION OF VEGETATION TYPES IN STATE PARK ARAGUAIA (MT)

The identification and discrimination of different vegetation types are essential activities for the study of impacts of human activities, one of the most viable alternatives for planning management actions, conservation and monitoring in protected areas (UC). The objective of this study was to evaluate the digital classifier Support Vector Machine (SVM) and the Spectral Angle Mapper (SAM) in the characterization of existing vegetation types in the Araguaia State Park (PEA), in the municipality of Novo Santo Antonio, being bounded by the confluence of the Death River and Araguaia River. Through the SVM classifiers and SAM were processed images ASTER (Advanced Thermal Emission and Reflection Spacebone Radiometer). The collection of training samples to generate the digital classifications comprehended the classes: water, mounds fields, Ipucas (forest fragments located in depressions that favors its flooding), monchao (high ground of island covered with savannah species), savanna, cerrado, shoals and bare soil. The samples provided the training for the two algorithms. In SVM algorithm were used three kernel options. In the SAM algorithm were used two alpha values (alpha). Based on the Kappa index was found that the SVM algorithm obtained more precise differentiation of vegetation types, obtaining greater accuracy in the mapping.