Research on Intelligent Extraction Methods for Key Regions in Hyperspectral Images

In response to the problem of high complexity and inaccurate extraction of key areas in hyperspectral satellite images, this paper proposes an intelligent extraction method for key areas in hyperspectral images. Firstly, the feature layer fusion target spectral signal matching detection method is adopted, and the subspace projection method is used to suppress background clutter signals and enhance the contrast between the target and background clutter to extract the target area. In the absence of any prior information on key area targets, this method uses anomaly detection algorithms to perform coarse screening on abnormal areas. Finally, key area detection is achieved through the fusion of target spectral signal matching detection and anomaly detection decision level fusion in the feature layer. Experimental results show that the proposed method can effectively suppress clutter signals, enhance the distinguishability between targets and backgrounds, and handle situations whether or not the prior spectral information is known. Moreover, the algorithm has strong robustness and low processing complexity.