基于HySime算法的端元个数估计  

Endmember Number Estimation Based on Hyperspectral Signal Identification by Minimum Error

作　　者：侯璨 李志飞 HOU Can;LI Zhifei(Hunan Branch of Shanghai Feiwei Information Technology Co.,Ltd.,Changsha 410005,China;Hunan Satellite Image Information Technology Co.,Ltd.,Changsha 410004,China)

机构地区：[1]上海飞未信息技术有限公司湖南分公司,湖南长沙410005 [2]湖南星图空间信息技术有限公司,湖南长沙410004

出　　处：《地理空间信息》2024年第4期92-95,共4页Geospatial Information

摘　　要：基于最小误差的高光谱信号识别(hyperspectral signal identification by minimum error,HySime)是一种估计高光谱端元个数的算法,该算法首先使用多元回归估计信号和噪声相关矩阵,然后使用信号相关矩阵的特征向量子集来表示信号子空间。为了科学评估HySime算法,分别对不同信噪比的高斯白噪声、高斯有色噪声模拟高光谱数据以及马蹄湾村真实高光谱数据的端元个数进行估计。实验表明HySime算法自适应性强,稳定性好,在解算过程中不需要输入任何参数,就能准确估计高光谱数据的端元个数。Hyperspectral signal identification by minimum error(HySime)is an algorithm for estimating the number of hyperspectral endmembers.The algorithm uses multiple regression to estimate the signal and noise correlation matrix,and then uses a eigenvector subset of signal correlation matrix to represent the signal subspace.In order to scientifically evaluate HySime algorithm,we estimated the endmember number of Gaussian white noise,Gaussian colored noise simulated hyperspectral data with different signal-to-noise ratios and the real hyperspectral data of Matiwan Village.Experimental results show that HySime algorithm has strong adaptability and good stability,does not need to input any parameters during the calculation process,and can accurately estimate the endmember number of hyperspectral data.

关 键 词：HySime算法 高光谱影像 混合像元分解 端元个数估计 

分 类 号：P237[天文地球—摄影测量与遥感]