Hyperspectral anomaly detection method based on multi-feature joint background reconstruction subtraction

The invention relates to an anomaly detection method for a hyperspectral image. According to the method, anomaly detection is carried out on the hyperspectral image based on multi-feature combined background reconstruction subtraction. The method specifically comprises the following steps: firstly, performing background reconstruction on an image by using a double-window trilateral filtering method jointly improved by spatial features and spectral features; secondly, for analyzing distribution characteristics of abnormal targets in the image, providing a saliency feature extraction method based on global context perception on the basis of a traditional saliency detection method to extract a saliency feature map of the image; secondly, performing difference square on the saliency characteristic graph of the image and the reconstructed background graph after trilateral filtering to obtain an abnormal target initial detection graph; and finally, obtaining a spectral weight graph of the image by using the spectra