一种高光谱遥感图像快速谱聚类算法  

A fast spectral clustering algorithm for hyperspectral remote sensing images

作　　者：张亚平[1] 张宇[1] 杨楠 罗晓[3] 罗谦[3] ZHANG Yaping;ZHANG Yu;YANG Nan;LUO Xiao;LUO Qian(School of Transportation Science and Engineering,Harbin Institute of Technology,Harbin 150090,China;Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Land and Resources,Shenzhen 518034,China;The Second Research Institute of CAAC,Chengdu 610041,China)

机构地区：[1]哈尔滨工业大学交通科学与工程学院,黑龙江哈尔滨150090 [2]国土资源部城市土地资源监测与仿真重点实验室,广东深圳518034 [3]中国民用航空局第二研究所,四川成都610041

出　　处：《测绘通报》2019年第12期60-64,共5页Bulletin of Surveying and Mapping

基　　金：国家自然科学基金重点基金(U1533203);国土资源部城市土地资源监测与仿真重点实验室开放基金(KF-2018-03-026)

摘　　要：为获得分类效果更优良的遥感图像分类方式并解决高光谱遥感图像分类运算速度缓慢的问题,集成Lanczos算法与谱聚类算法,探讨了高光谱遥感图像谱聚类算法应用于遥感图像分类的可行性,提出了一种面向高光谱遥感图像的快速谱聚类算法;通过对比美国圣地亚哥机场高光谱遥感图像K-均值算法与谱聚类算法的分类结果,发现面向高光谱遥感图像的谱聚类算法易于识别线性地物,且分类的速度能得到较大提升。In order to obtain a better classification method of remote sensing images and solve the problem of slow classification operation of hyperspectral remote sensing images,this paper integrates Lanczos algorithm and spectral clustering algorithm,discusses the feasibility of spectral clustering algorithm of hyperspectral remote sensing images applied to remote sensing image classification,and proposes a hyperspectral remote sensing image-oriented fast spectral clustering algorithm.By comparing the classification results of Kmeans algorithm and spectral clustering algorithm for hyperspectral remote sensing image of San Diego Airport,USA,it is found that spectral clustering algorithm for hyperspectral remote sensing image is easy to recognize linear features,and the classification speed can be greatly improved.

关 键 词：高光谱遥感图像 谱聚类算法 K-均值算法 LANCZOS算法 图像分类 

分 类 号：P237[天文地球—摄影测量与遥感]