Hyperspectral anomaly detection method based on superpixel guide discrimination forest

The invention provides a hyperspectral anomaly detection method based on a superpixel guide discriminant forest, and mainly solves the problems of poor local abnormal target detection performance and high false alarm rate in the existing method. According to the scheme, the method comprises the following steps: 1) extracting three most major component features of a hyperspectral image through a principal component analysis method, and performing multi-scale superpixel segmentation by using the features to obtain multi-scale superpixel features containing all wavebands; 2) constructing a discrimination forest model based on a plurality of band gain criteria to train and test the features of each super-pixel block, and generating initial detection images of a plurality of scales; and 3) performing fusion optimization on all the initial detection images through a multi-scale fusion model based on guide filtering to obtain a final detection result. According to the method, superpixel segmentation based on space a