Abstract: ‍ ‍Hyperspectral data has hundreds of bands with a huge amount of data. While providing rich and detailed information， there is strong correlation between different bands， especially between adjacent bands， leading to a large amount of information redundancy. To solve this problem， a joint Skewness-Kurtosis index band selection method based on cluster was proposed for hyperspectral anomaly detection. Firstly， the virtual dimension is used to estimate the intrinsic dimension of hyperspectral data， and then the clustering center is updated with the idea of maximum and minimum distance to avoid the problem that the initial values are randomly selected too close to each other. Then， considering that the abnormal target does not satisfy the background Gaussian distribution， the joint Skewness-Kurtosis index is used as the criterion function to select the important bands effectively. Experiments on three representative hyperspectral data sets show that the proposed algorithm can effectively improve the performance of hyperspectral anomaly detection and reduce the false alarm rate.