Abstract:This paper investigates some key technical points in the application of hyperspectral methods for small target detection. Starting from the data structure of hyperspectral images, it utilizes the high spectral and spatial correlations present in hyperspectral data. Simulations are conducted on the Botswana and Dioni datasets to verify the highly correlated characteristics of spectral and spatial data in hyperspectral imagery. After dimensionality reduction of the hyperspectral images using the PCA algorithm, image classification experiments are conducted. Additionally, the combination of the traditional RX target detection algorithm with the K-means clustering algorithm is studied. The simulation experiments demonstrate that the image classification after PCA processing and the algorithm combining RX with K-means not only improves detection accuracy but also significantly reduces detection time.