Band Selection Algorithm Of Hyperspectral Image Based On Clonal Selection Algorithm

Due to high spectral resolution,hyperspectral data carries a rich spectrum information, and can detect geophysics spectral characteristics of slight differences.Thus theoretically hyperspectral data has stronger ability to identify different object , which creat prerequisites for more accurately identify object, and it is also the superiority of hyperspectral imaging data . But at the same time hyperspectral data increases and high-dimensional hyperspectral data brings its treatment hardships, hyperspectral data "dimension disaster" problem appear, so it need to study more reasonable mathematical model to guide hyperspectral image analysis processing.Band selection is the best method to reduce the spectral dimension and solve hyperspectral dimension disasters effectively.Band selection means to choose the best band composition to build up new hyperspectral image space, it will not lose important information, and can reduce data dimension, and ensure the effective information extracted.In order to construct a algorithms for band selection in hyperspectral images which has both excellent behavior and low computation load ,this paper first introduces a algorithms for supervision band selection in hyperspectral images based on clonal selection algorithm, which has good effect and fast search function. Implication of supervision is that band selection process need to limit the number of band combinationt o fixed value.Band selection problem is equivalent to an optimization problem, so will the clonal selection algorithm for band selection could theoretically achieve very good effect. This band selection method will use clonal selection algorithm as search strategy, kappa coefficient as the band combination rule, in a relatively short period of time obtain satisfactory band combination. Variation is the main operation of clonal selection algorithm and plays a vital role, this paper proposes a new variation method based on spectral Angle of nonsymmetrical mutation, using AVIRIS data analyses to do simulation experiment. Experimental results show that variation based on spectral Angle of nonsymmetrical mutation clonal selection algorithm in re-ligion band performance has a very good improvement effect. Then, this paper also put band selection of hyperspectral images based on clonal selection algorithm in search speed and search performance and SFS ,band selection of hyperspectral images based on genetic algorithm were compared. Finally, this paper also puts forward a set of unsupervised band selection, unsupervised band selection in the supervision and band selection is conducted on the basis of the global o ptimal band combination search, can automatically determine band combination band number, output the final band combination; And choose the 20 dimension of hyperspectral data to test.