Hyperspectral Anomaly Detection Using Robust Principal Component Analysis with Autoencoding Adversarial Network

This paper introduces an unsupervised anomaly detection method for hyperspectral (HS) images, integrating robust principal component analysis (RPCA) and a deep autoencoding adversarial network (AEAN). The proposed method is solved using a plug-and-play alternating direction method of multipliers (PnP-ADMM). This method uses AEAN as a proximity operator to promote low-rankness in reconstructed spectral data, effectively distinguishing anomalous pixels from the background. Moreover, the performance of anomaly detection is enhanced by using the anomaly map estimated by the proposed method as weights for the WLRX detector. Experimental results show that our method outperforms some existing methods in detection accuracy.