High order statistics based anomaly detection research for hyperspectral remote sensing

Based on RX algorithm which proposed by Reed and Yu, a new anomaly detector was proposed. And a series of measures were proposed to process the anomaly points. SAM (spectral angle mapper) was used to judge the points which next to the anomaly points on some conditions, and the eligible points were considered as parts of anomaly target, then dilating the imagery. And the dilated imagery showed the anomaly targets which was detected by the detector. At the same time, SAM was used for the anomaly point classification. The value of SAM between two anomaly points vectors was not beyond the setting threshold, and the two anomaly points were considered as the same class. The basic analysis depended on the result of the classification. The performance of the new detector is proved by the PHI data experiment.