The spectral classifiers allow a good estimate for the mapping of the materials from the similarity between the reference curve and the image. Initially the spectral classifiers had been developed for hyperspectral images analysis. However, some works demonstrate good results for the application of these techniques in multispectral images. The present work aims to evaluate the spectral classifier Spectral Identification Method (SIM) in ASTER image. The Spectral Identification Method (SIM) is proposed to establish a new similarity index and three estimates according to the significance levels (5%, 10% and 15%) of the materials. This method is based on two statistical procedures: ANOVA and Spectral Correlation Mapper (SCM) coefficient. This information can be used to evaluate the degree of correlation among the materials in analysis. The advantage of this method is to validate according to levels of significance of the most probable areas of the sought material. The method was applied to ASTER image at the Military Instruction Field located Formosa (GO) close to Federal District. The images were acquired with atmosphere correction. The pixels size from the SWIR image were duplicated in order to join the VNIR and SWIR images. Endmembers were detected in three steps: a) spectral reduction by the Minimum Noise Fraction (MNF) transformation, b) spatial reduction by the Pixel Purity Index (PPI) and c) manual identification of the endmembers using the N-dimensional visualizer. The classification was made from the endmembers of nonphotosynthetic vegetation (NPV), photosynthetic vegetation (PV) and soil. These procedures allowed to identify the main scenarios in the study area.
remote sensing; image processing; spectral classifier
