Hyperspectral Anomaly Detection with Multivariate Skewed t Background Model

Abstract

In this paper, autoencoder-based multivariate skewed t-distribution is proposed for hyperspectral anomaly detection. In the proposed method, the reconstruction error between the hyperspectral images reconstructed by the autoencoder and the original hyperspectral images is calculated and is modeled with a multivariate skewed t-distribution. The parameters of the distribution are estimated using the variational Bayes approach, and a distribution-based rule is determined for anomaly detection. The experimental results show that the proposed method has better performance when compared to the RX, LRASR and DAEAD anomaly detection methods.