LAUSR

Hyperspectral anomaly detection (HAD) has always been a hot topic in hyperspectral image (HSI) processing. The Reed-Xiaoli (RX) detector is one of the most widely used methods for HAD. However, most improved versions of RX are usually only improved from the view of reducing the contamination of anomaly pixels on the background modeling. Considering the rich spatial information of HSI and the influence of anomaly pixels on background modeling, we propose an ensemble and random RX with multiple features (ERRX MFs) anomaly detector. First, the spectral, Gabor, extended morphological profile (EMP), and extended multiattribute profile (EMAP) four features are extracted. Then, random subsampling is used on RX to obtain multiple detection results for each feature. Finally, the detection results of these four features are further fused to the final detection result by ensemble learning. Experimental results on four real hyperspectral datasets validate that our proposed method can achieve more satisfactory results compared with other state-of-the-art methods.