基于人工蜂群算法高光谱图像波段选择

Artificial bee colony algorithm-based band selection for hyperspectral imagery

摘要:

为减少高光谱遥感图像光谱空间冗余、降低计算复杂度,提出一种基于人工蜂群算法的高光谱图像波段选择方法. 首先,根据波段相关性矩阵对全波段进行预处理,获得相关性较小的波段子空间;然后,利用人工蜂群算法以最佳指数与JM距离的加权和为适应度函数在各子空间进行邻域搜索,不断更新至收敛为止,从而获得最优波段组合. 最后,利用AVIRIS数据和ROSIS数据对提出的算法与基于蚁群,粒子群,拟态物理学算法的波段选择方法进行实验.仿真结果表明：基于人工蜂群算法的波段选择能够在保证良好收敛性的同时,大大降低计算花费,所获得的波段组合用于高光谱图像分类时,可以得到较好的分类精度.

Abstract:

A hyperspectral image band selection algorithm based on artificial bee colony algorithm isproposed to reducespectralredundancy of hyperspectral remote sensing image and computational complexity.Firstly, accordingtothecorrelationcoefficientmatrices among bands some pretreatments have been taken too btain the band sub space with less relevance. Then, neighborhood search has been implemented on each sub-space by using artificial bee colony algorithm together with the weighted sum between JM distance and OIF as the fitness function.To obtain the optimal band combination,the search is updated until the algorithm is convergent. Finally, the proposed algorithmis used to compare with band selection methods based on ACO, PSO and APO. The experimental results show that the proposed algorithm can not only ensure a good convergence but also reduce the computational cost. Simultaneously, when the obtained bands combination is used for hyperspectral image classification, higher classification accuracy can be obtained.