摘要
遥感图像波段选择是遥感数据应用的前提,它帮助人们对遥感图像进行可视化分析和解译,并且能够增强图像的质量,体现地表不同地物之间的差异性,为目标识别、图像分类、变化检测提供数据基础。但是,高光谱图像波段数目众多,即光谱维度高的特性给高光谱图像波段组合带来了巨大问题与挑战。所以,对高光谱数据进行降维处理是必要的。波段选择研究中,为了保持原始波段的光谱特性不变,特征(波段)选择方法是最合理的降维方法。在原始数据集合中选择特定波段构成波段子集,随后进行波段组合研究。本文设计了一种改进近邻子空间划分(IASP)的方法,构建了基于视觉显著性的波段选择模型在模型中通过对比各个典型显著性检测算法的效果,最终选择HC(histogram-based Contrast)显著性算法选择显著波段,并设计对比实验,利用珠海一号高光谱卫星数据,验证了该方法的有效性。
The selection of remote sensing image bands is a prerequisite for the application of remote sensing data.It aids in the visualization and interpretation of remote sensing images,enhances image quality,and highlights the differences between different surface features.This provides a foundational basis for target recognition,image classification,and change detection.However,the large number of hyperspectral image bands,that is,the high spectral dimension,brings great problems and challenges to the band combination of bloom images.Therefore,it is necessary to reduce the dimension of hyperspectral data.In research on band combination,to preserve the spectral characteristics of the original band,the feature selection method is the most reasonable dimensionality reduction approach.In the original data set,select a specific band to form a band subset,and then carry out band selection research.In this paper,an Improved adjacent subspace partition(IASP)method is designed,and a band selection model based on visual saliency is constructed.Finally,the Histogram-based Contrast algorithm is selected to select the significant band,and a Contrast experiment is designed to verify the effectiveness of the method using the data of the OrbitaHyperSpectral satellite.
作者
杨桄
胡昊文
金椿柏
任春颖
王龙光
王琪
刘文婧
YANG Guang;HU Hao-wen;JIN Chun-bai;REN Chun-ying;WANG Long-guang;WANG Qi;LIU Wen-jing(Aviation University of Air Force,Changchun 130022,China;Northeast Institute of Geography and Agroecology,Chinese Academy of Sciences,Changchun 130102,China)
出处
《光谱学与光谱分析》
北大核心
2025年第10期2950-2959,共10页
Spectroscopy and Spectral Analysis
基金
国家重点研发计划项目(2021YFD1500105)资助。
关键词
高光谱图像
视觉显著性
波段选择
波段组合
最佳指数因子
Hyperspectral image
Visual salience
Band selection
Band combination
Optimal index factor
作者简介
杨桄,1975年生,空军航空大学教授,e-mail:yg2599@126.com;通讯作者:任春颖,E-mail:renchy@iga.ac.cn;通讯作者:刘文婧,E-mail:Liuwenjing130@sina.com。
