Features extraction method based on intrinsic mode function for hyperspectral data

The empirical mode decomposition (EMD) theory was applied and the features extraction method based on intrinsic mode function (IMF) was proposed in order to eliminate the errors of parameters estimation(such as covariance matrix) and retain the detail spectral information. The maximum value, minimum value and mean of hyperspectral data were calculated to estimate the IMF. IMF can express the spectral absorption features with different scales of hyperspectral data. The raw hyperspectral data was projected the IMF dimension to implement the spectral features extraction of hyperspectral data. The airborne hyperspectral data collected by push-broom hyperspectral imager (PHI) was applied to analyze and evaluate the performance of the proposed method. The results show that the effect of covariance singularity and inaccurate parameters estimation of hyperspectral data is avoided, the main and important information of data is retained and the classes' separability is increased.