辽宁工程技术大学学报（自然科学版）2024，Vol.43Issue(3)：366-372,7.DOI:10.11956/j.issn.1008-0562.20240214
基于适应度函数和染色体信息量排序的高光谱影像特征选择方法
Hyperspectral image feature selection method based on fitness function and chromosome information quantity ranking
摘要
Abstract
To solve the problem of redundant information in hyperspectral remote sensing image data,a fitness function of joint conditional mutual information and multivariate mutual information is constructed based on the information ranking of chromosomes to improve the amount of information provided by the selected features.The fitness function is used as the evaluation standard of differential evolution algorithm,and the optimal feature subset is obtained by maximizing the fitness function.A new spectral feature selection algorithm is proposed.The correlation is calculated using the amount of information of the selected feature in each chromosome.The experimental results show that the algorithm achieves the maximum classification accuracy on 9 out of 16 categories of ground objects,indicating that the estimation based on the correlation of information content as the fitness function combined with the swarm intelligence optimization algorithm can be better applied to the spectral feature selection of hyperspectral remote sensing images.关键词
高光谱/差分进化算法/多元互信息/特征选择/适应度函数Key words
hyperspectral/differential evolution algorithm/mutual information/feature selection/fitness function分类
计算机与自动化引用本文复制引用
钱韫竹,吕欢欢..基于适应度函数和染色体信息量排序的高光谱影像特征选择方法[J].辽宁工程技术大学学报（自然科学版）,2024,43(3):366-372,7.基金项目
辽宁省自然科学基金项目(20180550450) （20180550450）
