Anomaly Detection in Hyperspectral Imagery Based on Spectral Dimensions Transformation and Spatial Filter

In this paper, a novel anomaly detection algorithm is proposed for hyperspectral imagery, which is the extended RX algorithm based on spectral dimension transformation and spatial filter (STSF-RX). Firstly, minimum noise fraction (MNF) transform is performed on the original hyperspectral images, by setting a SNR threshold, and obtains MNF transform matrices that the SNR of their corresponding bands are larger than the threshold. Then, for suppressing background interferences, orthogonal subspace projection (OSP) is estabished by MNF transform matrices, and project the hyperspectral data of MNF transform to the OSP, and obtain the error data of hyperspectral. In order to concentrate the energy of detection targets in first a few components, principal components analysis (PCA) method is performed on the processing hyperspectral image. Finally, we obtain bands for PCA transform based on the eigenvalue threshold, the eigenvalue of the bands are larger than threshold, and the bands are input to the RX detector. The simulation results demonstrate that the proposed STSF-RX algorithm outperforms the others algorithm, it is higher precision and lower false alarm probability; especially, the computation time of the proposed algorithm is very short.