Based on the clustering of the background for hyperspectral imaging anomaly detection

RX algorithm is the most classical algorithm in hyperspectral image anomaly detection algorithm, but the detection effect down significantly in a complicated and nonhomogeneous background. This paper use EM algorithm to smooth background by clustering the adjacent area of the pixel under test (PUT); in the process of detection, using the average of clustering replace the original background, in order to reduce the influence of the background complexity on the detection algorithm. With AVIRIS hyperspectral data, the simulation experiment has good detection effect.