A back-propagation neural network for mineralogical mapping from AVIRIS data

Imaging spectrometers have the potential to identify surface mineralogy bas ed on the unique absorption features in pixel spectra. A back-propagation n eural network (BPN) is introduced to classify Airborne Visible/Infrared Ima ging Spectrometer (AVIRIS) of the Cuprite mining district (Nevada) data int o mineral maps. The results are compared with the traditional acquired surf ace mineralogy maps from spectral angle mapping (SAM). There is no misclass ification for the training set in the case of BPN; however 17 percent miscl assification occurs in SAM. The validation accuracy of the SAM is 69 percen t, whereas BPN results in 86 per cent accuracy. The calibration accuracy of the BPN is higher than that of the SAM, suggesting that the training proce ss of BPN is better than that of the SAM. The high classification accuracy obtained with the BPN can be explained by: (1)its ability to deal with comp lex relationships (e.g., 40 dimensions) and (2) the nature of the dataset, the minerals are highly concentrated and they are mostly represented by pur e pixels. This paper demonstrates that BPN has superior classification abil ity when applied to imaging spectrometer data.