The high dimensionality of hyperspectral image increases information,but it also leads to the question of dimensionality curse.There are several problems to be solved in reducing dime nsion,eliminating redundancy of bands,and suppressing background interferences during hyperspectral anomaly targets detection.Aiming at the problems,this paper proposes a new anomaly target detection algorithm of hyperspectral image based on particle swarm optim ization (PSO) clustering.Firstly,the algorithm optimizes the traditional-means clustering by using PSO method,the original h yperspectral image is divided for bands subset class by PSO clustering while the features of hyperspectral image bands aren′t cha nged,and those bands with similar features are clustering;Then, the feature information of all band subsets is extracted by using the principa l component analysis,which makes the information of anomaly target with highlight and suppresses background interference;At last,the optimal band subsets are achieved by fourth-order cumulant of principa l component in band subsets,and anomaly de tection is carried on by the kernel RX.The results show that the proposed algorithm was higher precision and lower false alarm probability.