Lithology classification of L-band SAR data by combining backscattering and polarization characteristics
Taking Mindoro Island in Philippines as the study area,and selecting ALOS PALSAR dual-polariza-tion data(polarization mode of HH and HV polarization)as the source data,the authors used the maximum likeli-hood classification method to classify the rock units and geological mapping in study area by extracting the backscat-tering coefficients(Sigma0_HH and Sigma0_HV)and the polarization decomposition parameters(entropy,angle and anisotropy).With the inclusion of polarization decomposition parameters,the overall accuracy increased from 36.706％to 65.000％,comparing to the result using only backscattering coefficients.The F1 scores of the coastal marsh and coral reef exceeded 0.80,and those of the gabbro and Mansalay Formation exceeded 0.75.After adding polarization features,the rock unit boundaries were better extracted and increased separability of the Mansalay Forma-tion and Mindoro metamorphic rocks from other rocks.Three polarization decomposition parameters compensate for the similarity of backscattering coefficients of multiple rock units and significantly improve the separability of rock units.It is shown that the combination of polarization decomposition parameters and backscattering coefficients of L-band SAR data can improve the classification accuracy of rock units in vegetation-covered areas.
NETL
NSTL
万方数据
