Abstract: Based on the principles of band selection,in this paper,a new method of band selection,named ABO,was proposed by combing the subspace partition,the band selection method based on matrix mode(BSMM),and the optimum index factor(OIF).In ABO,firstly,based on the correlation of bands,all bands are divided into different subspaces.Then, the index “W” is gotten through the method of BSMM,and only one band is seclected in each subspace,in which the seclected band has the maximal “W”.After this,the OIF from each three bands of the seclected is calculated.At last,the new method is tested by the real hyperspectral remote sensing imagery,and the three seclected bands is carried on RGB synthesize,HSV transformation and the RX anomaly detection.The results show that the proposed method has the best effect.

Key words: Hyperspectral,Band selection,Correlation,Subspace