Print

出版日期: 2016-09-25
点击次数:
下载次数:
DOI: 10.11834/jrs.20166169
2016 | Volumn20 | Number 5





                              上一篇|





下一篇


论文
混合像元分解技术及其进展
expand article info 陈晋 , 马磊 , 陈学泓 , 饶玉晗
北京师范大学 地表过程与资源生态国家重点实验室, 北京 100875

摘要

混合像元分解模型是定量遥感研究的重要组成部分,为各种地学应用提供了更精细的亚像元级地物信息,这一领域受到国内外学者们广泛关注。本文围绕混合像元分解研究的4个核心问题——光谱混合模型、端元提取、模型反演方法以及解混精度评估,总结了近20年来混合像元分解的重要研究进展,分析和介绍了典型算法模型的原理和思路。进一步阐述了现有研究在一些关键问题上存在的不足,如目前仍缺乏公认的线性和非线性模型的选择判据、已有的混合像元分解模型无法抑制由端元光谱相关造成的共线性问题。最后总结了混合像元分解未来的发展趋势和值得探索的研究方向。如结合辐射传输模型和地面试验,定量分析多次散射的影响机制,以及结合克服共线性的统计回归模型。

关键词

线性混合像元分解 , 非线性混合模型 , 端元光谱变异 , 共线性 , 软分类精度

Research progress of spectral mixture analysis
expand article info CHEN Jin , MA Lei , CHEN Xuehong , RAO Yuhan
State Key Laboratory of Earth Surface Processes and Resource Ecology(ESPRE), Beijing Normal University, Beijing 100875, China

Abstract

Spectral Mixture Analysis (SMA) is one of the main topics in quantitative remote sensing research. It is able to provide land cover information at sub-pixel levels for practical applications. With the emergence of improved algorithms, SMA has made significant progress in many aspects, including spectral mixture models, endmember determination, endmember fraction inversion, and accuracy assessment. This study focused on these four key components in SMA and reviewed the available models and algorithms developed in last two decades. Moreover, the deficiencies of existing studies were analyzed. These deficiencies include the absences of widely accepted model selection criteria for linear and nonlinear spectral mixture analysis models and the unstable inversion of existing spectral mixture analysis caused by the high spectral correlation between endmembers. Finally, the study summarized the directions for future research, which include quantitatively evaluating the amplitude and spectral shape of multiple scattering among endmembers, identifying the factors that contribute to the nonlinear component in mixture observed signals by using radiative transfer models and laboratory measurement experiments, improving the robustness of linear spectral mixture analysis models, and suppressing high sensitivity to noise error signals resulting from collinearity with some insights from available statistical regression models for collinearity issues.

Key words

linear spectral mixture analysis , nonlinear spectral mixture analysis , multiple scattering , endmember spectral variability , collinear effect endmember purification , soft classification accuracy ,

1 引言

遥感影像以像元(pixel)为基本单位获取地物信息,所记录的是像元内所有地物辐射能量的总和。混合像元在遥感影像中普遍存在,不仅影响地物识别和分类的精度,也是遥感科学向定量化发展面临的主要挑战之一。

近几十年来,国内外研究提出了多种混合像元分解技术,其基本假设是:在观测像元内,地表由少数几种光谱特征稳定的地物(端元,endmember)组成,观测的混合反射率为端元的光谱和它们面积比例(丰度,abundance)的函数(Adams等,1986Keshava和Mustard,2002):

$ \boldsymbol{M} = G({f_1}{f_2} \ldots {f_n};{\boldsymbol{R}_1}{\boldsymbol{R}_2} \ldots {\boldsymbol{R}_n}) $ (1)

式中,M向量为混合光谱,fn表示第n类端元的丰度比例,Rn向量为第i类端元的光谱,G为光谱混合函数,主要包括线性模型和非线性模型两种。混合像元分解技术的关键步骤可概括为:(1)根据地物的辐射传输特性,确定光谱混合模型的数学表达形式;(2)通过地面实测或图像端元提取算法获取端元光谱;(3)在已知混合光谱及端元光谱前提下,通过特定的数学方法(如最小二乘法),反演混合像元中各端元的面积比例(盖度/丰度)。

本文将围绕混合像元分解中的模型选择、端元提取、反演方法以及精度评估4个关键问题,阐述国内外近20年的进展和存在的主要问题。

2 光谱混合模型及选择判据

由于地表光谱混合过程的复杂性,目前尚无统一的模型可以适用于各种类型的地表光谱混合过程。光谱混合模型主要分为线性模型和非线性模型两类。

2.1 线性混合像元分解模型

线性混合模型LMM(Linear Mixture Model)忽略地物间的多次散射(Adams等,1986),认为混合光谱是端元光谱及其丰度的线性组合:

$ \boldsymbol{M}{\rm{ = }}\sum\limits_{k = 1}^n {{f_k}{\boldsymbol{R}_k} + \boldsymbol{\varepsilon} } $ (2)

式中,M为该像元的混合反射率;fk为第k端元的丰度;Rk为端元反射率;n为混合像元内端元的总数;ε为残差项。该模型通常用最小二乘法进行求解得到端元丰度。为保证求解结果不失物理意义,常附加约束条件:端元丰度总和为1约束ASC(Abundance Sum-to-one Constraint)及非负性约束ANC(Abundance Non-negativity Constraint)(Shimabukuro和Smith,1991)。

$ \sum\limits_{k = 1}^n {{f_k}} {\text{ = }} 1, \;\; 0 \leqslant {f_k} \leqslant 1 $ (3)

线性混合模型由于物理含义明确、模型简单,得到了广泛的应用。

2.2 非线性混合像元分解模型

考虑到端元间的多次散射,学者们提出了多种非线性光谱混合模型,包括:几何光学模型(李小文和王锦地,1995)、概率模型(Ju等,2003)、神经网络模型(Foody等,1997)以及高次多项式模型(Ray和Murray,1996Fan等,2009Halimi等,2011Nascimento和Bioucas-Dias,2009Somers等,2009a)。其中,高次多项式模型的物理意义比较明确、形式简单,目前应用最为广泛(此后文中的非线性模型均指该模型)。

高次多项式模型(Ray和Murray,1996)是在线性模型基础上,引入交叉端元来表示多次散射,用于研究植被与土壤的混合场景。此后,Nascimento和Bioucas-Dias(2009)提出了不限于植被和土壤场景的双线性模型:

$ {\boldsymbol{M}}{\text{ = }}\sum\limits_{k = 1}^n {{f_k}{{\boldsymbol{R}}_k} + \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {{f_{ij}}{{\boldsymbol{R}}_i}{{\boldsymbol{R}}_j}} }+{\boldsymbol{\varepsilon}}} $ (4)

式中,RiRj为交叉端元,即第ij两端元的反射率的Harmond积,代表二次散射。fijij端元间二次散射作用的贡献比。为保证模型反演结果不失物理意义,非线性模型也可在ASC和ANC约束下进行:

$ \begin{gathered} \sum\limits_{k = 1}^n {{f_k} + \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {{f_{ij}}} } } {\text{ = }} 1, \hfill 0 \leqslant {f_k}, \; {f_{ij}} \leqslant 1 \hfill \end{gathered} $ (5)

Fan等人(2009)认为二次散射作用的贡献(即fij)与其所构造的端元的面积有关(即fifj),提出了另一种双线性模型:

$ {\boldsymbol{M}}{\text{ = }}\sum\limits_{k = 1}^n {{f_k}{{\boldsymbol{R}}_k} + \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {{f_i}{f_j}{{\boldsymbol{R}}_i}{{\boldsymbol{R}}_j}} }+ {\boldsymbol{\varepsilon}} } $ (6)

Halimi等人(2011)提出了广义双线性模型GBM(Generalized Bilinear Model),该模型是在交叉端元前添加了调节因子:

$ {\boldsymbol{M}}{\text{ = }}\sum\limits_{k = 1}^n {{f_k}{{\boldsymbol{R}}_k} + \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {{\gamma _{i,j}}{f_i}{f_j}{{\boldsymbol{R}}_i}{{\boldsymbol{R}}_j}} }+ {\boldsymbol{\varepsilon} }} $ (7)

当调节因子γij为1时,则GBM模型变为式(6);γij为0时,模型变为线性模型。

由于非线性模型较好地刻画了复杂场景的多次散射作用,在实际应用中解混精度往往优于线性模型(Chen和Vierling,2006Somers等,2009a)。

2.3 模型选择判据

在端元空间结构复杂的混合场景(植被与土壤混合)中,多次散射作用明显,非线性模型往往能更准确地反演端元丰度。然而,非线性模型包含交叉端元,使得共线性风险增加,解混误差也可能大于线性模型(Chen等,2011)。

对于信号混合复杂场景而言,如何选择合适的模型(线性或非线性)是关键问题。而模型选择的判据取决于混合信号的非线性程度,即多次散射项的比例。当多次散射不可忽略时,端元的线性组合无法解释非线性混合信号,此时应选择非线性模型。若场景中各种端元的立体结构简单时,光子在各种端元间发生二次碰撞的概率降低,线性模型表现较好(Elmore等,2000)。因此,通过实测或辐射传输模型模拟的方式,定量化混合像元中的多次散射强度,分析影响多次散射因素(如端元的面积比、空间分布、高度差异等)将有助于获得模型选择依据。Wang等人(2015)利用FLIGHT辐射传输模型和地面实验,定量评估了灌丛与土壤间的多次散射效应。研究发现,随灌木覆盖度增加,多次散射占混合信号的比例逐渐增加,近红外波段尤为明显,高达15%。因此对于植被盖度较高的地区,建议选用非线性模型。

现有的非线性模型多假设多次散射以二次散射为主,而更高次散射的信号能量微弱,可忽略不计,因此交叉端元中仅包括二次散射项(Heylen等,2014Somers等,2014)。但Wang等人(2015)指出,随着灌木盖度和高度的增加,三次及高次散射作用的比例迅速上升,当灌木面积大于80%或冠幅高于1 m时,三次及高次散射比例高于30%(Wang等,2015)。此时,交叉项应该包括三次及高次散射端元。

然而,即使模型仅包含二次散射,随着像元内端元数的增加,交叉端元个数也呈指数增加,模型变得极为复杂。事实上,并非所有端元间的二次散射作用均需要考虑,当非线性模型包含太多不参与光谱混合的端元时,反演的误差会变大。因此,应剔除不必要的二次散射效应,确定有效的端元组合。Somers等人(2009a)发现在果树、杂草、土壤3种端元的混合场景中,当非线性模型仅包含果树间、果树与杂草间的二次散射时,解混结果最优,残差最小。Raksuntorn和Du(2010)提出了可变端元的非线性混合像元模型,模型利用端元间的空间关系确定非线性模型的端元集。宋梅萍等人(2014)提出了基于有效端元集的双线性模型,将端元按照其与混合像元的光谱相似性排序,通过分析排序结果与误差变化情况,剔除未参与光谱混合的交叉端元。

尽管学者们在定量测量和模拟多次散射作用及其各个组分的贡献方面开展了研究,但线性和非线性模型的选择仍缺乏公认的判据,需要经验性地预估不同场景中的多次散射作用的形式和强度。

3 引言

3.1 端元提取算法

端元提取是混合像元分解的关键步骤,根据是否假设遥感图像中存在纯端元,可将现有算法大致分为两类。一是端元识别算法EEA(End-member Extraction Algorithms),主要包括N-FINDR法(Winter,2004)、像元纯度指数法PPI(Pixel Purity Index Boardman,1994)、序列最大角度凸锥法SMACC(Sequential Maximum Angle Convex Cone Gruninger等,2004)、顶点成分法VCA(Vertex Component Analysis Nascimento和Dias,2005)。这类算法假设:图像中存在各种地物的纯像元,可将所有混合像元看作是高维空间中的多面体,而多面体的顶点即为端元,而不同算法的主要区别在于利用不同的投影变换和搜索策略确定多面体的顶点。

第二类算法是端元生成算法EGA(Endmember Generation Algorithm),这类算法不依赖于图像存在纯像元这一假设,直接生成纯端元的光谱特征,目前有迭代误差分析法IEA(Iterative Error Analysis)(Sun等,2008)、极小体积变换法MVT (Minimum Volume Transform Craig,1994)、最小体积约束的非负矩阵分解法MVC-NMF(Minimum Volume Constraint Nonnegative Matrix Factorization Miao和Qi,2007)、基于光谱提纯的非线性混合像元分解模型EP-NSMA(Endmember Purification Nonlinear SMA Ma等,2015)。

这两大类方法相比而言,EGA更加接近现实情况,而EEA在一些情况下出现失效。如在干旱与半干旱地区,植被盖度较低,可能不存在植被纯像元。Plaza等人(2005)选用AVIRIS高光谱数据,对比了EEA和EGA算法的表现,发现EGA算法提取的与地面真实端元光谱的光谱角更接近。李二森等人(2011)对比了6种不同的EEA算法和EGA算法的端元提取精度和运算效率,发现EGA的提取精度优于EEA,但相对费时。

近年来学者们也提出将空间信息和端元识别相结合的提取算法。如Plaza等人(2002)提出基于数学形态学的端元提取算法AMEE(Automated Morphological Endmember Extraction);Rogge等人(2007)提出SSEE算法(Spatial-Spectral Endmember Extraction)。

3.2 端元光谱变异性

端元光谱变异(endmember variability)指同类端元光谱出现差异(Somers等,2011)。作为混合像元分解的重要误差来源,国内外学者提出的解决方法主要分为两类:

一类是基于光谱特征的方法,此类方法通过特征选取或光谱变换等手段减少端元内光谱变异性。主要包括:选择光谱变异性较弱的短波红外波段的方法(Asner和Lobell,2000)、稳定区分解方法SZU (Stable Zone Unmixing Somers等,2010),该方法设计了基于端元内变异性和端元间变异性的不稳定指数ISI(InStability Index),通过该指数筛选光谱变异性弱的特征波段。此外,Chang和Ji(2006)Somers等人(2009b)提出了加权混合像元分解方法WSMA(Weighted Spectral Mixture Analysis),通过将各个波段的端元光谱变异特征表达为权重,使得端元内变异性弱的特征对混合模型的作用得以强化。光谱特征变换方法有:Wu(2004)提出的归一化混合像元分解方法NSMA(Normalized Spectral Mixture Analysis);陈学泓等人(2009)提出了基于Fisher判别分析的混合像元分解方法;Zhang等人(2004)提出的基于微分变换的混合像元分解DSU(Derivative Spectral Unmixing)。考虑到微分变换倾向于扩大高频误差,Li(2004)用小波变换代替微分变换,提出了基于离散小波变换的混合像元分解技术(discrete wavelet based SMA)。Rivard等人(2008)提出了基于连续小波变换的混合像元分解技术(continuous wavelet based SMA)。总的来说,基于光谱特征方法计算复杂度较低,易实现及应用。

另一类是基于多端元思想的方法,认为像元中的端元数量、光谱及其组合可以不一致,通过迭代计算和概率论的思路可以将多种端元混合的可能性进行全面考虑,进而获得分解误差最小的组合作为最终的解混结果。Roberts等人(1998)最早提出多端元混合像元分解MESMA(Multiple Endmember Spectral Mixture Analysis)。与MESMA相似,丛浩等人(2006)提出了一种端元可变的混合像元分解方法,基于相关系数选择可能的端元组合进行解混。此外,Bateson等人(2000)提出了端元束方法EBM (Endmember Bundles Method),Song(2005)提出了贝叶斯混合像元分解BSMA(Bayesian Spectral Mixture Analysis),该方法用概率密度函数来描述同类端元中不同光谱出现的概率分布,进而根据贝叶斯理论推导出端元盖度的概率密度函数。总体而言,上述方法物理意义清晰,将端元内光谱变异性造成的所有可能端元组合都进行考虑,这可能使得每个像元所得到的端元组合都不一致,这一结果更加符合实际的复杂地表覆盖组合情况。但这类方法需要较大的计算成本,尤其是在地表异质性较强的区域。此外,单纯采用分解误差最小判定原则,也可能得到不合乎现实情况的端元组合。

4 模型反演及共线性问题

求解混合像元模型一般使用带约束的最小二乘法,即线性或非线性模型的残差达到最小:

$ \begin{gathered} \min (\boldsymbol{\varepsilon} {^{\rm{T}}}\boldsymbol{\varepsilon} ) = {\rm{min}}[{(\boldsymbol{M} - \boldsymbol{\widehat M})^{\rm{T}}}(\boldsymbol{M} - \boldsymbol{\widehat M})]\\ {\text{s.t.}} \left\{ {\begin{array}{*{20}{c}} {\sum\limits_{i = 1}^n {{f_i}} {\text{ = }} 1 {\text{或}} \sum\limits_{k = 1}^n {{f_k} + \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {{f_{ij}}} } } {\text{ = }} 1 } \\ {0 \leqslant {f_i}, {f_{ij}}\leqslant 1 } \end{array}} \right. \hfill \\ \end{gathered} $ (8)

式中,$\hat M$为线性模型或非线性模型拟合值,ε是线性模型或非线性模型的残差。

最小二乘模型要求模拟信号和真实信号在绝对值上尽可能接近,因此对噪声信号比较敏感。为克服这一缺陷,Chen等人(2009)提出了通用混合像元分解理论框架,该模型考虑到端元变异性可分为形状差异以及强度差异两种,模型选用(相关系数、光谱角和光谱信息散度等)作为模拟信号和真实混合信号的光谱相似度函数,将混合像元分解过程转换为非线性最优问题。实验表明该模型可以在一定程度上降低端元光谱强度差异对混合像元分解造成的影响,也能部分抑制大气或地形因素带来的误差。

此外,地理数据往往存在空间自相关性,这与现有模型中各混合像元信号相互独立的假设相矛盾。为此,Van Der Meer(1999)提出了3个准则以抑制残差信号的空间自相关性。另外,学者们还提出端元的丰度也存在空间相关性,这一特性可纳入到现有的混合像元分解模型中。Jia和Qian(2007)提出SSCBSS模型(Spectral and Spatial Complexity Blind Source Separation),认为端元光谱及其分布的统计特性具有一定的空间自相关性,可以将光谱复杂度和空间复杂度引入到盲信号分离模型中,为高光谱图像解混提供了新思路。其他学者(Castrodad等,2011Iordache等,2012)尝试在反演时,增加空间正则化算子。Song等人(2010)利用Moran’I指数,描述端元丰度的空间一致性,认为相似的混合像元的端元丰度也接近,模型在最小化残差信号的同时应最大化端元丰度的空间自相关性。Shi和Wang(2014)对结合空间信息的端元提取和解混方法与传统模型进行了全面对比,指出结合空间信息和光谱信息的混合像元分解模型是未来值得探索的方向。

混合像元中不同端元之间光谱可能具有一定程度的相似性,导致混合像元分解模型出现共线性问题(Gong和Zhang,1999Chen等,2011),模型估算的丰度对噪声非常敏感,反演严重失真。然而此前,只有部分学者(Chen等,2011; Van Der Meer和Jia,2012)注意到了混合像元分解中的共线性问题,并做了一定的讨论。Chen等人(2011)较充分地讨论了非线性光谱混合模型分解误差与端元共线性程度、随机噪声之间的关系,特别指出,在非线性模型中,由于交叉端元由真实端元光谱相乘而得,模型中产生共线性的可能性大大增加。Ma等人(2016)提出了两步约束的非线性模型TsC-NSMA(Two-step Constrained Nonlinear Spectral Analysis),以解决非线性模型的共线性问题,该模型可以较好地抑制非线性模型的共线性问题,但并未彻底解决该问题。

5 精度评估

在评价混合像元分解模型的解混精度时,现有研究多采用均方根误差RMSE(Root Mean Square Error)和平均偏差AD(Average Deviation)来度量模型反演的端元丰度与参考数据的接近程度。RMSE主要用于反映模型反演结果的绝对偏差,而AD则用于评价模型的统计偏差。

混合像元分解可看作是广义的软分类器,因此软分类的精度评价方法也适用于评价混合像元分解模型。其中亚像元混淆矩阵SCM(Sub-pixel Confusion Matrix)较为常用,它是在传统的基于硬分类的混淆矩阵的精度评价方法上进一步改进得到(Pontius和Cheuk,2006Silván-Cárdenas和Wang,2008)。利用SCM矩阵,可计算模型的整体精度OA(Overall Accuracy)、用户精度UA(User Accuracy)、生产者精度PA(Producer Accuracy)以及Kappa系数。

Chen等人(2010)利用模拟影像,评价了RMSE与基于SCM的4个评价指标的一致性。结果表明:当检验样本足够多,各个端元类别的误差可比时,RMSE、OA、UA、PA均可用于评价模型。但检验样本有限或各种端元间误差的分布有明显差异时,基于SCM的OA和Kappa更适合用于混合像元分解模型评价。

6 结语

尽管混合像元分解在线性和非线性模型、端元提取算法、克服端元变异性的方法、模型反演以及精度评估等方面取得一定进展,但其中有一些关键问题亟待解决:

(1)以实验测量及辐射传输模型模拟为基础,定量化多次散射效应的强度以及其影响因素,提出简单的多次散射效应估算模型,为线性和非线性模型的选择提供物理判据。

(2)考虑端元光谱变异的两种形式,利用描述光谱“形状”和“幅度”特征的匹配算子作为模型丰度反演的目标函数,以克服不同端元的光谱变异。此外,在克服光谱变异的同时,还需考虑其他误差因素如大气校正误差、多次散射、传感器噪声等多种误差来源的叠合影响。

(3)光谱混合模型的共线性问题使得模型的丰度估计严重失真,但统计学家提出的处理共线性问题的方法尚未在混合像元分解中得到充分应用,如何结合新的统计方法和混合像元分解的特点,发展克服混合像元分解中共线性问题的有效方法将是未来研究的重点方向之一。

参考文献(References)

  • Adams J B, Smith M O, Johnson P E.1986.Spectral mixture modeling:a new analysis of rock and soil types at the Viking Lander 1 Site. Journal of Geophysical Research:Solid Earth, 91 (B8): 8098–8112. DOI: 10.1029/JB091iB08p08098.
  • Asner G P, Lobell D B.2000.A Biogeophysical approach for automated SWIR Unmixing of soils and vegetation. Remote Sensing of Environment, 74 (1): 99–112. DOI: 10.1016/S0034-4257(00)00126-7.
  • Bateson C A, Asner G P, Wessman C A.2000.Endmember bundles:a new approach to incorporating endmember variability into spectral mixture analysis. IEEE Transactions on Geoscience and Remote Sensing, 38 (2): 1083–1094. DOI: 10.1109/36.841987.
  • Boardman J W.1994.Automating Spectral Unmixing of AVIRIS Data Using Convex Geometry Concepts.
  • Castrodad A, Xing Z M, Greer J B, Bosch E, Carin L, Sapiro G.2011.Learning discriminative sparse representations for modeling, source separation, and mapping of hyperspectral imagery. IEEE Transactions on Geoscience and Remote Sensing, 49 (11): 4263–4281. DOI: 10.1109/TGRS.2011.2163822.
  • Chang C I, Ji B H.2006.Weighted abundance-constrained linear spectral mixture analysis. IEEE Transactions on Geoscience and Remote Sensing, 44 (2): 378–388. DOI: 10.1109/TGRS.2005.861408.
  • Chen J, Jia X P, Yang W, Matsushita B.2009.Generalization of subpixel analysis for hyperspectral data with flexibility in spectral similarity measures. IEEE Transactions on Geoscience and Remote Sensing, 47 (7): 2165–2171. DOI: 10.1109/TGRS.2008.2011432.
  • Chen J, Zhu X L, Imura H, Chen X H.2010.Consistency of accuracy assessment indices for soft classification:simulation analysis. ISPRS Journal of Photogrammetry and Remote Sensing, 65 (2): 156–164. DOI: 10.1016/j.isprsjprs.2009.10.003.
  • Chen X H, Chen J, Jia X P, Somers B, Wu J, Coppin P.2011.A quantitative analysis of virtual endmembers' increased impact on the collinearity effect in spectral Unmixing. IEEE Transactions on Geoscience and Remote Sensing, 49 (8): 2945–2956. DOI: 10.1109/TGRS.2011.2121073.
  • Chen X X, Vierling L.2006.Spectral mixture analyses of hyperspectral data acquired using a tethered balloon. Remote Sensing of Environment, 103 (3): 338–350. DOI: 10.1016/j.rse.2005.05.023.
  • Chen X H, Wang S Q, Chen J, Shen M G, Zhu X L.2009.New algorithm for spectral mixture analysis based on fisher discriminant analysis:evidence from laboratory experiment. Journal of Infrared and Millimeter Waves, 28 (6): 476–480. DOI: 10.3321/j.issn:1001-9014.2009.06.017. ( 陈学泓, 王胜强, 陈晋, 沈妙根, 朱孝林. 2009. 一种新的基于Fisher判别的混合像元分解算法:室内控制实验结果分析. 红外与毫米波学报, 28 (6): 476–480. DOI: 10.3321/j.issn:1001-9014.2009.06.017. )
  • Cong H, Zhang L P, Li P X.2006.A Method of Selective Endmember For Pixel Unmixing. Journal of Image and Graphicsv, 11 (8): 1092–1096. DOI: 10.11834/jig.200608185. ( 丛浩, 张良培, 李平湘. 2006. 一种端元可变的混合像元分解方法. 中国图象图形学报, 11 (8): 1092–1096. DOI: 10.11834/jig.200608185. )
  • Craig M D.1994.Minimum-volume transforms for remotely sensed data. IEEE Transactions on Geoscience and Remote Sensing, 32 (3): 542–552. DOI: 10.1109/36.297973.
  • Elmore A J, Mustard J F, Manning S J, Lobell D B.2000.Quantifying vegetation change in semiarid environments:precision and accuracy of spectral mixture analysis and the normalized difference vegetation index. Remote Sensing of Environment, 73 (1): 87–102. DOI: 10.1016/S0034-4257(00)00100-0.
  • Fan W Y, Hu B X, Miller J, Li M Z.2009.Comparative study between a new nonlinear model and common linear model for analysing laboratory simulated-forest hyperspectral data. International Journal of Remote Sensing, 30 (11): 2951–2962. DOI: 10.1080/01431160802558659.
  • Foody G M, Lucas R M, Curran P J, Honzak M.1997.Non-linear mixture modelling without end-members using an artificial neural network. International Journal of Remote Sensing, 18 (4): 937–953. DOI: 10.1080/014311697218845.
  • Gong P, Zhang A.1999.Noise effect on linear spectral unmixing. Geographic Information Sciences, 5 (1): 52–57. DOI: 10.1080/10824009909480514.
  • Gruninger J H, Ratkowski A J and Hoke M L.2004.The sequential maximum angle convex cone (SMACC) endmember model//Defense and Security:International Society for Optics and Photonics.Baltimore:SPIE:1-14
  • Halimi A, Altmann Y, Dobigeon N, Tourneret J Y.2011.Nonlinear Unmixing of hyperspectral images using a generalized bilinear model. IEEE Transactions on Geoscience and Remote Sensing, 49 (11): 4153–4162. DOI: 10.1109/TGRS.2010.2098414.
  • Heylen R, Parente M, Gader P.2014.A review of nonlinear hyperspectral Unmixing methods. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7 (6): 1844–1868. DOI: 10.1109/JSTARS.2014.2320576.
  • Iordache M D, Bioucas-Dias J M, Plaza A.2012.Total variation spatial regularization for sparse hyperspectralUnmixing. IEEE Transactions on Geoscience and Remote Sensing, 50 (11): 4484–4502. DOI: 10.1109/TGRS.2012.2191590.
  • Jia S, Qian Y T.2007.Spectral and spatial complexity-based hyperspectralUnmixing. IEEE Transactions on Geoscience and Remote Sensing, 45 (12): 3867–3879. DOI: 10.1109/TGRS.2007.898443.
  • Ju J C, Kolaczyk E D, Gopal S.2003.Gaussian mixture discriminant analysis and sub-pixel land cover characterization in remote sensing. Remote Sensing of Environment, 84 (4): 550–560. DOI: 10.1016/S0034-4257(02)00172-4.
  • Keshava N, Mustard J F.2002.Spectral unmixing. IEEE Signal Processing Magazine, 19 (1): 44–57. DOI: 10.1109/79.974727.
  • Li E S, Shu S L, Zhou X M, Yu W J.2011.The development and comparison of endmember extraction algorithms using hyperspectral imagery. Journal of Remote Sensing, 15 (4): 659–679. DOI: 10.11834/jrs.20110156. ( 李二森, 朱述龙, 周晓明, 余文杰. 2011. 高光谱图像端元提取算法研究进展与比较. 遥感学报, 15 (4): 659–679. DOI: 10.11834/jrs.20110156. )
  • Li J.2004.Wavelet-based feature extraction for improved endmember abundance estimation in linear unmixing of hyperspectral signals. IEEE Transactions on Geoscience and Remote Sensing, 42 (3): 644–649. DOI: 10.1109/TGRS.2003.822750.
  • Li X W, Wang J D. Optical remote sensing model and structure parameters of vegetation. Beijing: Science press 1995 . ( 李小文, 王锦地. 1995. 植被光学遥感模型与植被结构参数化. 北京: 科学出版社 . )
  • Ma L, Zhou Y, Chen J, Cao X, Chen X H.2015.Estimation of fractional vegetation coverin semiarid areas by integrating endmember reflectance purification into nonlinear spectral mixture analysis. IEEE Geoscience and Remote Sensing Letters, 12 (6): 1175–1179. DOI: 10.1109/LGRS.2014.2385816.
  • Ma L, Chen J, Zhou Y, Chen X H.2016.Two-step constrained nonlinear spectral mixture analysis method for mitigating the collinearity effect. IEEE Transactions on Geoscience and Remote Sensing, 54 (5): 2873–2886. DOI: 10.1109/TGRS.2015.2506725.
  • Miao L D, Qi H R.2007.Endmember extraction from highly mixed data using minimum volume constrained nonnegative matrix factorization. IEEE Transactions on Geoscience and Remote Sensing, 45 (3): 765–777. DOI: 10.1109/TGRS.2006.888466.
  • Nascimento J M P, Dias J M B.2005.Vertex component analysis:a fast algorithm to Unmixhyperspectral data. IEEE Transactions on Geoscience and Remote Sensing, 43 (4): 898–910. DOI: 10.1109/TGRS.2005.844293.
  • Nascimento J M P and Bioucas-Dias J M.2009.Nonlinear mixture model for hyperspectral unmixing//Proceedings of SPIE 7477, Image and Signal Processing for Remote Sensing XV.Berlin, Germany:SPIE [DOI:10.1117/12.830492].
  • Plaza A, Martínez P, Pérez R, Plaza J.2002.Spatial/spectral endmember extraction by multidimensional morphological operations. IEEE Transactions on Geoscience and Remote Sensing, 40 (9): 2025–2041. DOI: 10.1109/TGRS.2002.802494.
  • Plaza A, Sánchez-Testal J J, Plaza J and Valencia D.2005.An experimental evaluation of endmember generation algorithms//Proceedings of SPIE 5995, Chemical and Biological Standoff Detection Ⅲ.Boston, MA:SPIE [DOI:10.1117/12.630778].
  • Pontius R G Jr, Cheuk M L.2006.A generalized cross-tabulation matrix to compare soft-classified maps at multiple resolutions. International Journal of Geographical Information Science, 20 (1): 1–30. DOI: 10.1080/13658810500391024.
  • Raksuntorn N, Du Q.2010.Nonlinear spectral mixture analysis for hyperspectral imagery in an unknown environment. IEEE Geoscience and Remote Sensing Letters, 7 (4): 836–840. DOI: 10.1109/LGRS.2010.2049334.
  • Ray T W, Murray B C.1996.Nonlinear spectral mixing in desert vegetation. Remote Sensing of Environment, 55 (1): 59–64. DOI: 10.1016/0034-4257(95)00171-9.
  • Rivard B, Feng J, Gallie A, Sanchez-Azofeifa A.2008.Continuous wavelets for the improved use of spectral libraries and hyperspectral data. Remote Sensing of Environment, 112 (6): 2850–2862. DOI: 10.1016/j.rse.2008.01.016.
  • Roberts D A, Gardner M, Church R, Ustin S, Scheer G, Green R O.1998.Mapping chaparral in the Santa Monica Mountains using multiple endmember spectral mixture models. Remote Sensing of Environment, 65 (3): 267–279. DOI: 10.1016/S0034-4257(98)00037-6.
  • Rogge D M, Rivard B, Zhang J, Sanchez A, Harris J, Feng J.2007.Integration of spatial-spectral information for the improved extraction of endmembers. Remote Sensing of Environment, 110 (3): 287–303. DOI: 10.1016/j.rse.2007.02.019.
  • Shi C, Wang L.2014.Incorporating spatial information in spectral Unmixing:a review. Remote Sensing of Environment, 149 : 70–87. DOI: 10.1016/j.rse.2014.03.034.
  • Shimabukuro Y E, Smith J A.1991.The least-squares mixing models to generate fraction images derived from remote sensing multispectral data. IEEE Transactions on Geoscience and Remote Sensing, 29 (1): 16–20. DOI: 10.1109/36.103288.
  • Silván-Cárdenas J L, Wang L.2008.Sub-pixel confusion-uncertainty matrix for assessing soft classifications. Remote Sensing of Environment, 112 (3): 1081–1095. DOI: 10.1016/j.rse.2007.07.017.
  • Somers B, Cools K, Delalieux S, Stuckens J, Van der Zande D, Verstraeten W W, Coppin P.2009a.Nonlinear hyperspectral mixture analysis for tree cover estimates in orchards. Remote Sensing of Environment, 113 (6): 1183–1193. DOI: 10.1016/j.rse.2009.02.003.
  • Somers B, Delalieux S, Stuckens J, Verstraeten W W, Coppin P.2009b.A weighted linear spectral mixture analysis approach to address endmember variability in agricultural production systems. International Journal of Remote Sensing, 30 (1): 139–147. DOI: 10.1080/01431160802304625.
  • Somers B, Verbesselt J, Ampe E M, Sims N, Verstraeten W W, Coppin P.2010.Spectral mixture analysis to monitor defoliation in mixed-aged Eucalyptus globulus Labill plantations in southern Australia using Landsat 5-TM, EO-1 Hyperion data. International Journal of Applied Earth Observation and Geoinformation, 12 (4): 270–277. DOI: 10.1016/j.jag.2010.03.005.
  • Somers B, Asner G P, Tits L, Coppin P.2011.Endmember variability in spectral mixture analysis:a review. Remote Sensing of Environment, 115 (7): 1603–1616. DOI: 10.1016/j.rse.2011.03.003.
  • Somers B, Tits L, Coppin P.2014.Quantifying nonlinear spectral mixing in vegetated areas:computer simulation model validation and first results. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7 (6): 1956–1965. DOI: 10.1109/JSTARS.2013.2289989.
  • Song C H.2005.Spectral mixture analysis for subpixel vegetation fractions in the urban environment:how to incorporate endmember variability?. Remote Sensing of Environment, 95 (2): 248–263. DOI: 10.1016/j.rse.2005.01.002.
  • Song M P, Zhang Y R, An J B, Bao H M.2014.Effective Endmember Based Bilinear Unmixing Model. Spectroscopy and Spectral Analysis, 34 (1): 196–200. DOI: 10.3964/j.issn.1000-0593(2014)01-0196-05. ( 宋梅萍, 张甬荣, 安居白, 包海默. 2014. 基于有效端元集的双线性解混模型. 光谱学与光谱分析, 34 (1): 196–200. DOI: 10.3964/j.issn.1000-0593(2014)01-0196-05. )
  • Song X F, Jiang X G and Rui X P.2010.Spectral unmixing using linear unmixing under spatial autocorrelation constraints//Proceedings of IEEE International Geoscience and Remote Sensing Symposium.Honolulu, HI:IEEE:975-978 [DOI:10.1109/IGARSS.2010.5649735].
  • Sun L X, Zhang Y and Guindon B.2008.Improved iterative error analysis for endmember extraction from hyperspectral imagery//Proceedings of SPIE 7086, Imaging Spectrometry XⅢ.San Diego, California, USA:SPIE [DOI:10.1117/12.799232].
  • Van Der Meer F.1999.Iterative spectral unmixing (ISU). International Journal of Remote Sensing, 20 (17): 3431–3436. DOI: 10.1080/014311699211462.
  • Van Der Meer F D, Jia X P.2012.Collinearity and orthogonality of endmembers in linear spectral unmixing. International Journal of Applied Earth Observation and Geoinformation, 18 : 491–503. DOI: 10.1016/j.jag.2011.10.004.
  • Wang J M, Cao X, Chen J, Jia X P.2015.Assessment of multiple scattering in the reflectance of semiarid shrublands. IEEE Transactions on Geoscience and Remote Sensing, 53 (9): 4910–4921. DOI: 10.1109/TGRS.2015.2413409.
  • Winter M E.2004.A proof of the N-FINDR algorithm for the automated detection of endmembers in a hyperspectral image//Proceedings of SPIE 5425, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery X.Orlando, FL:SPIE [DOI:10.1117/12.542854].
  • Wu C S.2004.Normalized spectral mixture analysis for monitoring urban composition using ETM+imagery. Remote Sensing of Environment, 93 (4): 480–492. DOI: 10.1016/j.rse.2004.08.003.
  • Xiao Q, Wen J G, Liu Q H, Zhou Y.2006.Study on spectral unmxing model and it's application in extracting chlorophyll concentration of water body. Journal of Remote Sensing, 10 (4): 559–567. DOI: 10.11834/jrs.20060482. ( 肖青, 闻建光, 柳钦火, 周艺. 2006. 混合光谱分解模型提取水体叶绿素含量的研究. 遥感学报, 10 (4): 559–567. DOI: 10.11834/jrs.20060482. )
  • Zhang J K, Rivard B, Sanchez-Azofeifa A.2004.Derivative spectral unmixing of hyperspectral data applied to mixtures of lichen and rock. IEEE Transactions on Geoscience and Remote Sensing, 42 (9): 1934–1940. DOI: 10.1109/TGRS.2004.832239.