The influence of spectral resolution on discriminating Brazilian sugarcane varieties

The influence of spectral resolution on discriminating five important Brazilian sugarcane varieties was studied using hyperspectral Hyperion data and the simulated response of some multispectral sensors. Discriminant analysis was applied to each dataset to enhance cultivar differentiation. Results showed that the variety SP80-1842 was easily discriminated in all datasets by a near-infrared band threshold. Discrimination between the other four cultivars with similar reflectance at each sensor dataset required discriminant analysis. The overall classification accuracy using a separate set of pixels presented the following values: AVHRR/3/NOAA-17 (58%), CCD/CBERS-2 (62%), HRG/SPOT-5 (66%), ETM+/LANDSAT-7 (72%), ASTER/TERRA (72%), MODIS/TERRA (74%), and Hyperion/EO-1 (87%). Discrimination of sugarcane varieties with the Hyperion instrument was confirmed by the comparison between ground data and the classification images derived from the discriminant analysis.