空间与谱间相关性分析的NMF高光谱解混

非负矩阵分解(NMF)技术是高光谱像元解混领域的研究热点。为了充分利用高光谱图像中丰富的空间与光谱相关性特征,改善基于NMF的高光谱解混算法性能,提出一种结合了空间与谱间相关性分析的NMF解混算法。算法针对NMF的通用性和局部极小问题,引入并结合高光谱图像两种典型的相关性特征,具体包括:基于马尔可夫随机场(MRF)模型,建立描述相邻像元空间相关特征的约束;通过复杂度映射技术,建立描述相邻波段谱间相关(光谱分段平滑)特征的约束;并将上述两种约束同时引入NMF解混目标函数中。实验结果表明,对于一般自然地物场景或人造地物场景,相对于分段平滑和稀疏约束的非负矩阵分解(PSNMFSC)、交互投影子梯度的非负矩阵分解(APSNMF)和最小体积约束的非负矩阵分解(MVCNMF)这3种代表性NMF解混参考算法,该算法可进一步提高高光谱解混精度;对于空间相关或谱间相关特征中某一种不显著的特殊场景,也具有更好的适应能力。通过将空间相关和谱间相关特征相结合,较全面地反映了高光谱数据与解混相关的重要特征,能够对绝大多数真实高光谱数据进行高精度解混,对高光谱解混及后续应用领域相关研究均具有参考价值。     

基于混合像元空间与谱间相关性模型的NMF线性盲解混

基于相关性分析的高光谱解混算法，通常缺少对高光谱图像空间和光谱相关性特征的综合分析与利用，或对于先验知识的依赖程度较高。本文提出一种基于混合像元空间与谱间相关性模型的NMF线性盲解混算法。具体包括：通过改进马尔科夫随机场（MRF）模型，建立相邻像元间的空间相关模型；利用复杂度映射技术，建立相邻波段间的光谱相关模型；在NMF目标函数外部和内部分别引入上述两种模型，作为盲解混算法的约束条件。试验结果表明，该算法相对于区域相关的NMF解混算法（ACBNMF）、最小化光谱相关度约束的NMF方法（MSCCNMF）和最小体积约束的非负矩阵分解（MVCNMF）等代表性NMF解混参考算法，解混精度有所提高；同时，降低了对于先验知识的依赖程度，拓宽了适用范围。     

采用目标端元修正的高光谱混合像元盲分解

针对非负矩阵盲信号分离(NMF)用于混合像元分解易陷入局部极小值的不足,将非监督端元提取与盲分解方法相结合,构建了一种基于目标端元修正的混合像元盲分解模型(ATGP-NMF)。ATGP-NMF模型利用非监督正交子空间投影算法(ATGP)和非负最小二乘法(NNLS)获取NMF盲分离的初始值,然后将获得初始目标端元光谱与丰度输入NMF模型,通过迭代运算不断逼近优化目标而得到最终的端元光谱和端元丰度。为了检验模型对于各类数据的有效性和适用性,将ATGP-NMF与传统NMF分别应用于模拟仿真数据、室内控制数据和真实遥感影像3类实验数据进行分析验证。结果表明,ATGP-NMF模型具有较好的适用性,在没有先验信息、先验信息很少,以及纯像元假设不存在情况下都能较好地分解混合像元,且能够更好克服局部极小问题,提高混合像元分解的精度。     

改进的高光谱解混初始化方法及其适用性分析

高光谱影像中存在大量的混合像元,极大地限制了高光谱影像的定量应用,高效且精准地进行像元解混尤为重要。端元矩阵的初始化、算法本身的代价函数及其迭代规则,三者的不同往往会导致获取的最终端元光谱和端元丰度的不同。在不同条件下,选取适当的初始化方法、代价函数和迭代规则,使得高光谱解混结果更优尤为重要。本文改进了一种基于欧氏距离和光谱信息散度的分块初始化方法(IBISS),改进后方法在中低信噪比情况下优于其他初始化方法。同时针对初始化、算法本身这两个方面进行大量试验,结果表明:①分块初始化优于全局初始化;②梯度迭代NMF算法相比于乘性迭代NMF算法,具有更快的收敛速度,但容易陷入局部最小值;③乘性迭代分块NMF算法相比于乘性迭代标准NMF算法能够获取更好的端元丰度信息;④梯度迭代分块NMF算法不适用于随机初始化后的光谱解混过程。     

线性混合光谱模型高光谱压缩感知

高光谱压缩感知(HCS)对于解决机载或星载高光谱数据的存储与实时传输具有重要意义。目前,线性混合模型(LMM)已被成功应用于HCS;然而,由于光照条件、地形变化以及大气作用等的影响,所获取的地物光谱会发生扰动,从而限制了HCS重建质量的提高。在LMM基础上,通过引入光谱修正项来修正光谱扰动,提出了光谱扰动修正的LMM (SPC_LMM);在此基础上,进一步提出了基于SPC_LMM的HCS (HCS_SPC_LMM)方法。该方法在采样端仅对原始高光谱图像进行光谱维压缩采样,基于压缩采样数据,将SPC_LMM应用HCS的重建,利用交替方向乘子法(ADMM)分别估计SPC_LMM中各分量的最优值,以获得最优的高光谱图像重建质量。实验结果表明,HCS_SPC_LMM能够获得优于其他典型HCS方法的重建质量。     

FAN模型的双种群差分进化光谱解混算法

双线性混合模型是近年来非线性光谱解混的研究重点之一,其克服了线性混合模型无法描述地物多重散射作用的缺陷,能够更精确地还原真实的地物光谱混合过程。然而,限于模型的复杂性,目前在缺乏准确的端元先验知识的条件下进行双线性光谱解混仍是一项具有挑战性的任务。差分进化算法(DE)是一种具有良好全局搜索能力的群智能优化算法,其优化求解过程无需进行复杂的数学推导,为双线性光谱解混问题提供了一种有效的解决途径。为此,本文以FAN双线性混合模型为例,提出了一种双种群机制的差分进化算法(记为DEFAN),实现非监督双线性光谱解混。DE-FAN算法通过建立端元与丰度两个种群的交替进化机制寻找最优解,同时在迭代中引入自适应重构策略增强种群多样性,降低算法陷入局部最优解的风险,最终实现端元与丰度的同时估计。通过模拟图像及真实图像的解混实验进行算法检验,证明DE-FAN算法较之传统非线性解混算法具有更高的解混精度及解混效率。     

基于单次散射反照率的矿物高光谱稀疏解混

矿物的混合多属于致密型混合,在可见光—短波红外波段的混合呈现非线性特征,同时由于矿物混合的复杂性以及图像中完全纯净的像元可能不存在等原因,使得从图像上提取端元具有较大不确定性。本文根据矿物单次散射反照率的线性可加性,提出一种基于矿物单次散射反照率光谱库的稀疏解混算法,利用Hapke模型将矿物反射率转换成矿物单次散射反照率,构建矿物单次散射反照率光谱库,以半监督的方式通过稀疏回归的方法从光谱库中寻找最优端元组合,并估算混合像元中各端元的丰度。利用RELAB矿物混合光谱库进行算法验证,结果表明,丰度反演的平均绝对误差为3.12%;将本文方法应用于美国内华达州铜矿区的AVIRIS高光谱图像数据,所得丰度图与美国地质勘探局USGS矿物识别结果具有较好的一致性。本文算法不需要从图像提取端元,并且考虑到了矿物的非线性混合特征,能够得到较高的反演精度,在近地行星和卫星表面岩矿成分的探测等领域具有较好的应用前景。     

基于高光谱Hyperion数据的线性光谱混合模型与神经网络模型的比较

混合像元问题是定量遥感中的热点问题之一,为了改进从遥感数据中提取定量信息,人们建立了各种混合光谱分解技术,其中线性光谱混合模型和神经网络模型就是两种比较成熟的方法。以陕西省横山地区的高光谱Hyperion数据为研究基础,通过最小噪声变换(MNF)、像元纯度指数(PPI)转换和RMS误差分析的迭代方法相结合提取影像中的纯净像元作为终端端元。分别运用神经网络模型和线性光谱混合模型对影像进行光谱分解,得到各个组分的分解图像。以标准植被指数(NDVI)影像为衡量标准,选取训练样本点,分别对两种模型进行回归分析,结果显示NDVI影像与线性光谱混合模型植被分解图像的判定系数(R2=0.91)要大于其与神经网络模型的判定系数(R2=0.81)。进一步分析表明在一般情况下,线性光谱混合模型具有比神经网络模型略高的分离精度,但是神经网络模型对细部信息的提取的效果要好于线性光谱混合模型,最后提出了端元均方根误差(EAR)指数,一种新的混合像元分解的思路。     

Scaling dimensions in spectroscopy of soil and vegetation

The paper revises and clarifies definitions of the term scale and scaling conversions for imaging spectroscopy of soil and vegetation. We demonstrate a new four-dimensional scale concept that includes not only spatial but also the spectral, directional and temporal components. Three scaling remote sensing techniques are reviewed: (1) radiative transfer, (2) spectral (un)mixing, and (3) data fusion. Relevant case studies are given in the context of their up- and/or down-scaling abilities over the soil/vegetation surfaces and a multi-source approach is proposed for their integration.Radiative transfer (RT) models are described to show their capacity for spatial, spectral up-scaling, and directional down-scaling within a heterogeneous environment. Spectral information and spectral derivatives, like vegetation indices (e.g. TCARI/OSAVI), can be scaled and even tested by their means. Radiative transfer of an experimental Norway spruce (Picea abies (L.) Karst.) research plot in the Czech Republic was simulated by the Discrete Anisotropic Radiative Transfer (DART) model to prove relevance of the correct object optical properties scaled up to image data at two different spatial resolutions. Interconnection of the successive modelling levels in vegetation is shown. A future development in measurement and simulation of the leaf directional spectral properties is discussed.We describe linear and/or non-linear spectral mixing techniques and unmixing methods that demonstrate spatial down-scaling. Relevance of proper selection or acquisition of the spectral endmembers using spectral libraries, field measurements, and pure pixels of the hyperspectral image is highlighted. An extensive list of advanced unmixing techniques, a particular example of unmixing a reflective optics system imaging spectrometer (ROSIS) image from Spain, and examples of other mixture applications give insight into the present status of scaling capabilities.Simultaneous spatial and temporal down-scaling by means of a data fusion technique is described. A demonstrative example is given for the moderate resolution imaging spectroradiometer (MODIS) and LANDSAT Thematic Mapper (TM) data from Brazil. Corresponding spectral bands of both sensors were fused via a pyramidal wavelet transform in Fourier space. New spectral and temporal information of the resultant image can be used for thematic classification or qualitative mapping.All three described scaling techniques can be integrated as the relevant methodological steps within a complex multi-source approach. We present this concept of combining numerous optical remote sensing data and methods to generate inputs for ecosystem process models.     

高光谱数据截断加权核范数稀疏解混

受仪器和观测条件限制,高光谱数据易受噪声污染,给数据解译带来挑战。针对传统稀疏解混模型抗噪性能差的问题,本文提出一种截断加权核范数稀疏解混方法,利用高光谱图像像元之间的相关性减轻噪声对丰度估计的干扰。该方法借助低秩表示在挖掘数据内在低维结构方面的优势,在稀疏解混中加入基于截断加权核范数的低秩约束,并结合加权稀疏技术,在稀疏正则项中引入空间邻域权重。截断加权核范数对丰度矩阵的奇异值向量分段处理,可以更好地实现丰度矩阵的低秩逼近,使丰度图像保持空间一致性并保留更多细节信息,空间加权策略则增强了丰度图像的空间连续性。模拟高光谱数据、Cuprite矿区真实数据和红树林高光谱数据实验表明,与其他先进的稀疏解混方法相比,所提方法具有更好的抗噪性,能够提高解混精度。     

Wheat crop inventory using high spectral resolution irs-p3 mos-b spectrometer data

Modular Optoelectronic Scanner (MOS-B) spectrometer data over parts of Northern India was evaluated for wheat crop monitoring involving (a) sub pixel wheat fractional area estimation using spectral unmixing approach and (b) growth assessment by red edge shift at different phenological stages. Red shift of 10 nm was observed between crown root initiation stage to flowering stage. Wheat fraction estimates using linear spectral unmixing on Feb. 13, 1999 acquisition of MOS-B data had high correlation (0.82) with estimates from Wide Field Sensor (WiFS) data acquired on same date by IRS-P3 platform. It was observed that five bands (4,5,8,12,13 MOS-B bands) are sufficient for signature separability of major land cover classes viz. wheat, urban, wasteland, and water based on purely spectral separability criterion using Transformed Divergence (T.D.) approach. Higher number of bands saturated the T.D. values. In contrast, performance of sub pixel fractional area estimation using unmixing decreased drastically for eight bands (4,5,6,7,8,9,12,13 MOS-B bands) chosen from optimal band selection criteria in comparison to full set of 13 bands. The relative deviation between area estimated from Wifs and MOS-B increased from 1.72 percent when all thirteen bands were used in unmixing to 26.10 percent for the above eight bands.     

火星表面含水矿物探测进展

火星表面含水矿物类型识别和空间分布特征研究对圈定火星表面生命活动有利区域和探索可能存在的火星生命形式具有重要科学意义。本文总结了20世纪90年代以来火星表面含水矿物的探测进展,从火星轨道器光谱仪遥感探测、着陆器和巡视器就位探测两方面介绍了矿物探测使用的数据源,重点阐述了目前火星表面已经探测到的各类含水硅酸盐矿物、硫酸盐矿物、碳酸盐矿物、氯盐及高氯酸盐矿物等含水矿物的光谱特征、矿物具体类别及分布特征,分析了火星表面含水矿物定量反演的主要方法与地质意义。最后从比较行星学角度倡议开展地球和火星含水矿物形成环境和形成过程的类比研究。     

Hyperspectral Pixel Unmixing via Spectral Band Selection and DC-Insensitive Singular Value Decomposition

This letter proposes a linear two-class hyperspectral pixel-unmixing algorithm that uses a band selection method to determine the best bands for pixel unmixing, low-pass prefiltering to remove high-frequency content, and a new version of the well-known singular value decomposition (SVD) method, which is insensitive to dc offsets (DCI-SVD). The proposed method is compared to the best level discrete wavelet transform approach for dimensionality reduction and least squares estimation and quadratic programming for unmixing (DWT-LSE-QP). The contributions of this letter are given as follows: (1) the band selection and filter selection algorithm and (2) the DCI-SVD algorithm. The dc insensitivity of the DCI-SVD method is proven, and simulation results using data from an analytical spectral device spectroradiometer show the efficacy of the proposed method and its superiority to the DWT-LSE-QP-based approach in the harder unmixing cases.     

南方典型植被遥感丰度信息提取

针对遥感影像在南方丘陵地区典型植被丰度信息提取中存在的大量混合像元问题，为进一步提高线性解混精度，通过计算像元EVI值，构建了Landsat 8时间序列影像南方典型植被（端元）和混合像元的EVI时间序列曲线，分析了不同生育期内各种地物类型的植被指数变化曲线，发现不同地物在植被指数时间序列中具有各自独立的波动规律。选取多个端元及其EVI时间序列曲线，利用光谱匹配方法对匹配EVI时间序列曲线和多个端元进行了匹配，达到利用不同端元组合进行光谱解混的目的。试验结果表明，与传统方法相比，阔叶林解混精度有明显提高，针叶林解混精度也有所提高。该研究成果可以为南方丘陵地区植被环境的研究提供有力支撑。     

Independent Base Vector Representation to Address Endmember Variability in Hyperspectral Unmixing

The normal compositional model (NCM) is a well-known and powerful model in hyperspectral unmixing which represents endmembers as independent Gaussian vectors to capture endmember variability. However, the assumption of independent endmembers diminishes the model accuracy because the high degree of correlation between endmembers of a scene and identical sources of variability demonstrate that the endmembers are dependent. This paper proposes a new hyperspectral unmixing algorithm which represents endmembers using dependent Gaussian vectors to estimate abundance fractions. To overcome the higher complexity caused by dependence assumption, this algorithm introduces new independent Gaussian vectors named Base Vectors to represent different endmembers by a weighted linear combination. Also, the proposed unmixing algorithm uses maximum likelihood method to estimate weight coefficients of Base Vectors which are used to represent mixed pixel. Finally, abundance estimation can be done using the new representation for endmembers and mixed pixel. The proposed algorithm is evaluated and compared with other state-of-the-art unmixing algorithms using simulated and real hyperspectral images. Experimental results demonstrate that the proposed unmixing algorithm can unmix pixels composed of correlated endmembers in hyperspectral images in the presence of spectral variability more accurately than previous methods.     

光谱解混下耕地土壤铬含量GMDH模型反演

针对遥感影像反射率与重金属元素间的光谱响应弱,土壤重金属经典反演模型精度较低等问题,本文以Sentinel-2号遥感影像为数据源,利用像元二分模型进行影像光谱解混,筛选出相关性较高的特征光谱作为光谱参量,构建基于像元线性解混和不同光谱变换下土壤反射率与重金属Cr含量的PLS模型和GMDH模型。研究结果表明,解混后的光谱与重金属Cr含量间的显著相关波段数增多,相关性增强。基于解混后的土壤光谱与重金属Cr含量构建的GMDH模型,其模型稳定性较好,预测能力更强,精度更好。该方法拓展了传统的利用遥感影像进行反演的思路,可为大范围监测土壤重金属的污染状况提供有益参考。     

基于支撑向量机概率输出的高光谱影像混合像元分解

提出利用支撑向量机（SVM）后验概率来分解高光谱影像的混合像元，通过支撑向量机的输出值转化为两两配对的后验概率，再由两两配对的概率值求得多类后验概率，并以像元所属类别的后验概率作为地物的组分信息。实验结果表明，该方法能较好地估计出混合像元的组分比。     

基于高光谱遥感数据提取森林结构参数的研究

以位于三峡库区的龙门河森林自然保护区为研究区,综合利用线性光谱混合模型和几何光学模型,基于高光谱遥感数据提取森林结构参数是本文研究的重点。在研究区地面调查数据的基础上,通过高光谱数据和混合光谱分解法,获得反演几何光学模型所需的四分量参数,根据背景光照分量与森林植被冠层各参数间的关系,反演得到森林冠层郁闭度及平均冠幅的定量分布图,并利用37个野外实测样本进行结果验证。     

基于线性混合模型的端元提取方法综述

混合像元是遥感领域研究的热点,而基于线性混合模型的光谱解混合技术正在越来越广泛地应用在光谱数据分析和遥感地物量化中,这项技术的关键就在于确定端元光谱。本文归纳了目前几种比较成熟的端元提取算法,分析了它们的主要思想和存在的优缺点,最后介绍了端元提取技术的应用及其发展趋势。     

Soil Organic Carbon mapping of partially vegetated agricultural fields with imaging spectroscopy

Soil Organic Carbon (SOC) is one of the key soil properties, but the large spatial variation makes continuous mapping a complex task. Imaging spectroscopy has proven to be an useful technique for mapping of soil properties, but the applicability decreases rapidly when fields are partially covered with vegetation. In this paper we show that with only a few percent fractional maize cover the accuracy of a Partial Least Square Regression (PLSR) based SOC prediction model drops dramatically. However, this problem can be solved with the use of spectral unmixing techniques. First, the fractional maize cover is determined with linear spectral unmixing, taking the illumination and observation angles into account. In a next step the influence of maize is filtered out from the spectral signal by a new procedure termed Residual Spectral Unmixing (RSU). The residual soil spectra resulting from this procedure are used for mapping of SOC using PLSR, which could be done with accuracies comparable to studies performed on bare soil surfaces (Root Mean Standard Error of Calibration = 1.34 g/kg and Root Mean Standard Error of Prediction = 1.65 g/kg). With the presented RSU approach it is possible to filter out the influence of maize from the mixed spectra, and the residual soil spectra contain enough information for mapping of the SOC distribution within agricultural fields. This can improve the applicability of airborne imaging spectroscopy for soil studies in temperate climates, since the use of the RSU approach can extend the flight-window which is often constrained by the presence of vegetation.