Feature extraction for hyperspectral remote sensing image using weighted PCA-ICA

Principal component analysis (PCA) and independent component analysis (ICA) are linear feature extraction methods in terms of the second-order statistics and higher-order statistics and have good compatibility and complementarity. For the feature extraction of the hyperspectral remote sensing image, an approach of the combined PCA and ICA was followed in the real remote sensing classification applications. In this study, the weighted PCA-ICA method was introduced to extract the feature information from HJ-1A hyperspectral imager (HSI) image. And then the real airborne visible infrared imaging spectrometer (AVIRIS) image case was performed by the distance similarity measure. Experimental results on HJ-1A HSI and AVIRIS images indicate that the proposed method can get high average accuracy of 89.55% and kappa coefficient of 0.8101 than the typical methods under certain condition with a suitable number of eigenvectors and weighted values.