Original spectral features contain information pertinent to certain target spectral features. Without an efficient spectral feature extraction method, the target detection performance might be degraded. We present spectral feature extraction techniques based on the Fourier domain for use in target detection. These feature extraction methods are the Fourier magnitude (FM), Fourier phase (FP), and Fourier coefficient selection (FCS) methods. In our target detection experiments, we compared the proposed methods to the principle component analysis (PCA) and independent component analysis (ICA) methods and the original spectral features. The experiment results show that the FCS target detection accuracy is 95.75%, whereas the accuracies of the FM, FP, PCA, ICA methods, and the original spectral features are 86.76%, 36.28%, 84.51%, 74.49%, and 78.92%, respectively. The average feature extraction times of the proposed methods are 223% faster than that found for the PCA and 304% faster than the ICA methods.
In this paper, we propose a method to reduce spectral dimension based on the phase of integrated bispectrum.
Because of the excellent and robust information extracted from the bispectrum, the proposed method can achieve
high spectral classification accuracy even with low dimensional feature. The classification accuracy of bispectrum
with one dimensional feature is 98.8%, whereas those of principle component analysis (PCA) and independent
component analysis (ICA) are 41.2% and 63.9%, respectively. The unsupervised segmentation accuracy of
bispectrum is also 20% and 40% greater than those of PCA and ICA, respectively.