The aim of this work was to assess Hyperion/EO-1 hyperspectral data to discriminate agricultural fields, comparing that data classification accuracy with ETM+/Landsat-7 multi-spectral data classification accuracy. For this purpose, agricultural targets with well-defined and subtle spectral differences in Franca municipality, São Paulo State, Brazil, imaged by both sensors in July, 2002, were discriminated by using a supervised classification (Maximum Likelihood) algorithm. Agricultural targets with well-defined spectral differences were characterized by six land use/land cover classes, while five sugarcane varieties were used as targets with subtle spectral differences. When the broadband ETM+ data were classified, the overall accuracy was 91.5% for six land use/land cover classes and 67.6% for five sugarcane variety classes, while for narrowband Hyperion data the accuracies were 94.9% and 87.1%, respectively. This result shows the importance of hyperspectral data use for agricultural targets discrimination with similar spectral characteristics.
digital classification; hyperspectral data; multispectral data
