Anomaly detection for hypaerspectral imagery using analytical fusion and RX

Anomaly detection is attractive for the analysis of hyperpectral imagery. This paper describes an expanded anomaly detection algorithm for small targets in hyperspectral imagery. As a variant of the well known multivariate anomaly detector called RX algorithm, the approach called the dimension reduction RX algorithm (DRRX) is proposed. The analytical fusion strategy is incorporated into the RX algorithm to lead to the DRRX algprithm. Experimental results are presented for the proposed DRRX and the classical constant false alarm rate (CFAR) RX algorithm for detecting small anomalies in hyperspectral imagery. The results show that the proposed DRRX algorithm outperforms the classical RX for detecting small targets in hyperspectral imagery.