地理空间信息2025，Vol.23Issue(4)：25-29,5.DOI:10.3969/j.issn.1672-4623.2025.04.006
基于多特征随机最近正则化子空间的高光谱影像分类算法
Hyperspectral Image Classification Algorithm Based on Multi-feature and Random Subspace Nearest Regularized Subspace
摘要
Abstract
The random subspaces nearest regularized subspace(RSNRS)based on a single feature has a limited accuracy,we proposed a new hy-perspectral remote sensing algorithm named multi-feature random subspace nearest regularized subspace(MFRSNRS).Firstly,we extracted Ga-bor features,LBP features,EMAPs features and spectral features by hyperspectral remote sensing images.Then,we selected the nearest regular-ized subspace(NRS)as the base classifier,and used the random subspace method to construct the feature subspace.Finally,we trained the base classifiers based on the feature subspace,and integrated the results of base classifiers to obtain the final classification results.The OA of MFRSNRS algorithm reached 92.88％in the Indian Pines data set,which is 17.05％higher than RSNRS algorithm.The OA of MFRSNRS algo-rithm reached 78.88％in the Pavia University data set,which is 7.73％higher than RSNRS algorithm.The experimental result indicates that the use of multiple features has shown a certain improvement in performance.关键词
多特征/最近规则子空间/高光谱影像分类/随机子空间Key words
multi-feature/NRS/hyperspectral image classification/random subspace分类
天文与地球科学引用本文复制引用
李倩楠,彭菲,虞瑶..基于多特征随机最近正则化子空间的高光谱影像分类算法[J].地理空间信息,2025,23(4):25-29,5.基金项目
江苏自然资源智库2023年度开放合作项目(ZK23002). （ZK23002）
