基于WOA和DPR的高光谱遥感影像分类算法

Classification Algorithm of Hyperspectral Remote Sensing Image Based on WOA and DPR

为保留原始波段特征的同时选择最有效的特征子集以得到理想的分类效果,提出一种鲸鱼优化算法和不连续保持松弛策略的组合优化算法.该算法首先以最大熵为适应度函数,对经过波段子空间划分的高光谱遥感影像使用鲸鱼优化算法进行最优波段子集的选择,将最优子集采用不连续保持松弛策略进行平滑.为验证该组合优化方法的有效性,使用高光谱遥感分类中经典的Indian Pines数据集和Pavia U数据集,进行分类精度评价.实验结果表明组合优化算法与其他传统分类方法相比具有更高的分类精度.

In order to select the most effective feature subset to obtain the ideal classification result while retaining the original band features,a combined optimization algorithm of whale optimization algorithm and discontinuity preserving relaxation(DPR)strategy was proposed in this paper.This algorithm firstly used the whale optimization algorithm(WOA)to select the optimal subset of bands for hyperspectral remote sensing images after the wave segment subspace partition,with the maximum entropy criterion(ME)as fitness function.Then discontinuity preserving relaxation method was used to smooth the optimal subset.To verify the effectiveness of the proposed combined optimization method,the classic Indian Pines and Pavia U data sets in the field of hyperspectral remote sensing classification are used to evaluate the classification accuracy.Experimental results show that the combined optimization algorithm proposed in this paper has higher classification accuracy than other traditional classification methods.