无线电工程2025，Vol.55Issue(5)：984-992,9.DOI:10.3969/j.issn.1003-3106.2025.05.010
基于超像素与纹理特征的高光谱图像分类方法
Hyperspectral Image Classification Method Based on Superpixels and Texture Features
摘要
Abstract
To solve the problems of low accuracy of classification results caused by insufficient sample numbers and low classification efficiency resulting from the extraction of redundant texture features in the classification of Hyperspectral Images(HSI),a HSI classification method based on superpixels and texture features is proposed.After using Principal Component Analysis(PCA)to extract the first six principal components of HSI as spectral bands,the six spectral bands are taken as the input for superpixel segmentation,and the point-based labeled samples are mapped into the superpixels.At the same time,the four texture feature factors with the largest contribution rates of the superpixel blocks in the directions of 0°,90°,180°,and 270° are obtained.By combining the texture features,spectral features,and spatial features,and based on the superpixel blocks,the Support Vector Machines(SVM)is used to complete the HSI classification.Experimental results show that the accuracy of the classification results on the two datasets is improved from 97.82％and 94.58％to 98.5％and 95.33％.In addition,compared with the scheme of extracting multiple texture features in multiple directions,the proposed method has better classification efficiency.关键词
高光谱/超像素/纹理特征/主成分分析/空-谱融合Key words
hyperspectral/superpixel/texture features/PCA/space-spectrum fusion分类
测绘与仪器引用本文复制引用
陈如俊,王子佳,张敏,曹帅帅,王雪峰,顾艳霜..基于超像素与纹理特征的高光谱图像分类方法[J].无线电工程,2025,55(5):984-992,9.基金项目
云南省教育厅科学研究基金项目(2024J2047)Scientific Research Fund Project of Yunnan Provincial Department of Education(2024J2047) （2024J2047）
