收稿日期: 2017-03-20
基金项目: 国家自然科学基金(编号:41671360)
第一作者简介: 张霞(1972— ),女,研究员,研究方向为高光谱遥感机理与信息提取。E-mail:zhangxia@radi.ac.cn
通讯作者简介: 吴兴(1993— ),男,硕士研究生,研究方向为火星表面含水矿物反演。E-mail:wuxing15@mails.ucas.ac.cn
中图分类号: TP751
文献标识码: A
|
摘要
矿物的种类与含量可限定其形成时特定的地质环境,因此火星表面矿物识别和丰度反演研究对了解火星的地质构造及历史演变过程具有重要意义。Eberswalde撞击坑具有复杂的水文系统,是火星探测的热点地区,而对该地区大范围矿物反演则鲜有研究。基于紧凑型侦查成像光谱仪(CRISM)目标探测模式数据和CRISM光谱库,首先,通过高光谱数据子空间识别算法(Hysime)分析CRISM数据信号子空间,并将光谱库所有矿物光谱投影到子空间上,保留投影误差较小的光谱作为约简后的端元光谱库;然后,使用协同稀疏解混对Eberswalde撞击坑西北部三角洲地区的矿物丰度进行定量反演,得到5种原生矿物:辉石、橄榄石、斜长石、菱铁矿、硬水铝石和1种蚀变矿物透闪石,并从矿物光谱、分布分别对反演结果进行验证;最后,从矿物学的角度对可能成因做出解释,推测该区域存在接触变质作用。
关键词
火星, Eberswalde撞击坑, 矿物, 协同稀疏分解, 丰度反演
Abstract
The type and content of minerals can constrain the specifically geological environment where they form. Therefore, studying the identification and abundance retrieval of minerals on Martian surface is essential in understanding the geological structure and historical evolution of Mars. Eberswalde Crater, a hotspot of Mars exploration, possesses a complex hydrological system. However, the large-scale mineral retrieval in this region has rarely been studied. Hence, we used hyperspectral technology to quantitatively retrieve the mineral abundances in the delta region of Eberswalde. Furthermore, we analyzed the geological environment that possibly existed at that time on the basis of the distribution and content of minerals. This study adopted a sparse unmixing algorithm to quantitatively retrieve mineral abundances by using Targeted Reduced Data Records (TRDR), which are obtained by Compact Reconnaissance Imaging Spectrometer for Mars (CRISM), together with the CRISM spectral library. However, the high mutual coherence of spectral libraries limits the performance of sparse unmixing because the equation to be solved is ill conditioned and time consuming. Therefore, we used the hyperspectral signal identification by minimum error (Hysime) algorithm to analyze the signal subspace of CRISM data. All the mineral spectra of the spectral library were projected onto the subspace. The spectrum with small projection error was then retained as the endmember, which truly contributed to the observed mixtures. Finally, a collaborative sparse unmixing algorithm was applied to CRISM to retrieve the mineral abundances of the delta region in Eberswalde. The abundance maps show five types of primary minerals (pyroxene, olivine, plagioclase, siderite, diaspore) and one type of alteration minerals (tremolite). Pyroxene is mainly distributed in front of the alluvial fan. Olivine is distributed (east-west) in the central area. Plagioclase is mainly distributed in the western delta region and a small crater edge in the northeastern corner. Siderite, with relatively less content, is mainly distributed along the edge of the alluvial fan. Diaspore is mainly distributed along the river valley, and the relatively enriched diaspore in the northeastern corner was probably caused by the transportation of liquid water in the past. Tremolite is distributed in the northeastern corner of the delta region. We identified a similar trend in both curves by comparing mineral spectra obtained from images with correspondent spectra in the spectral library. The result of the unmixing was consistent with the mineral distribution in the spectral index map, as shown by the comparison of the abundance maps of the three main minerals, namely, plagioclase, pyroxene, and olivine, and the correspondent spectral index map, thereby showing the effectiveness of the inversion results in this study. A collaborative sparse unmixing algorithm was used to quantitatively retrieve the mineral abundances of the delta region in Eberswalde. Six types of minerals were found, namely, pyroxene, olivine, plagioclase, siderite, diaspore, and tremolite. From the perspective of mineralogy, the instability of pyroxene and olivine indicated that the area where they are distributed is close to provenance and the original provenance is ultrabasic rock (e.g., peridotite) and basic rock (e.g., gabbro), respectively. The existence of tremolite and siderite indicated metasomatism between the rich water and the carbon dioxide fluid in this area. The minerals in the area of the alluvial fan were distributed outside the alluvial fan by fluid transportation. The presence of tremolite in the northeastern corner of delta region reflected the existence of contact metamorphism before the alluvial fan.
Key words
Mars, eberswalde crater, minerals, collaborative sparse unmixing, abundance retrival
1 引 言
火星表面以地貌或矿物的形式记录了火星早期水环境对地壳的改变,矿物的种类与含量可揭示其形成时特定的温度、压力、氧化还原环境等信息。研究火星矿物识别、空间分布及丰度含量,对于圈定过去液态环境范围(Mustard 等,2009)和理解火星的地质演化(Bibring 等,2006)具有重要意义。
高光谱遥感是地质岩矿信息探测最为有效的技术手段。火星表面矿物常用遥感探测器有热发射光谱仪(TES)(Bandfield,2002)、可见光及红外矿物填图光谱仪(OMEGA)(Bibring 等,2006;Ody 等,2012)及火星紧凑型侦查成像光谱仪(CRISM)(Murchie 等,2007),其中CRISM具有高光谱、高空间分辨率及较高的信噪比的特点,在火星矿物填图方面得到广泛的应用。目前火星矿物光谱研究主要有两种途径:(1)基于矿物诊断性光谱特征的光谱参数方法。Pelkey等人(2007)和Viviano-Beck等人(2014)根据CRISM传感器设计了一系列光谱参数产品(如波段比值、吸收深度等)判断特定类型矿物存在与否;Bishop等人(2013)基于CRISM数据根据光谱参数探测出层状硅酸盐在Mawrth Vallis地区广泛分布;Carter等人(2013)提出弱矿物信息自动识别方法,在CRISM数据提取含水硅酸盐矿物,并发现绿帘石。(2)基于混合像元分解的方法,如多端元线性光谱分解(Combe 等,2008)、独立组分分析(Moussaoui 等,2008)、非负矩阵分解等。Schmidt等人(2011)基于OMEGA数据运用迭代线性光谱解混、全约束最小二乘、全约束尺度梯度方法、基追踪去噪4种线性模型对Syrtis Major地区进行矿物解混;Poulet等人(2014)基于辐射传输模型Shkuratov,使用CRISM数据分别对好奇号4个备选着陆地点的一些出露点进行了矿物丰度反演。
CRISM光谱参数产品有利于对大区域尺度上矿物分布进行快速定性评估(Pelkey 等,2007),但实现岩矿信息的定量反演需借助于光谱分解获得像元内地物成分所占比例(丰度)。由于影像中很可能不存在纯净像元,端元选取将会限制解混精度。稀疏分解可避免端元提取的步骤,以半监督的方式从光谱库中寻找最优端元组合,来求解混合像元中各端元的丰度(林红磊 等,2016)。一个混合像元中实际存在的端元数量往往远小于光谱库中光谱的个数,因此光谱解混实际上是一个基于稀疏(即ℓ1范数)正则化的稀疏回归问题(Bioucas-Dias和Figueiredo,2010;Iordache 等,2011)。Bioucas-Dias和Figueiredo(2010)基于交叉方向乘子法(ADMM)发展了变量分裂和增广拉格朗日的稀疏解混算法(SUnSAL),之后又有学者对此算法做出改进(Iordache 等,2012a, 2014a, 2014b;Tang 等,2015)。Lin等人(Lin 等,2016;Lin和Zhang,2017)基于CRISM数据采用稀疏解混模型反演了Gale撞击坑好奇号着陆点附近区域含水矿物的丰度,为火星表面矿物定量反演提供了一种新的思路。
Eberswalde撞击坑地区具有复杂的水文系统,存在三角洲与河流冲刷渠道(Malin和Edgett,2003;Lewis和Aharonson,2006)。Poulet等人(2014)曾利用CRISM数据对Eberswalde三角洲的121个像元的平均光谱进行矿物反演,而大范围矿物反演对揭示过去的液态环境更有意义。故本文利用CRISM高光谱数据(http://ode.rsl.wustl.edu/mars/[2017-03-20])和光谱库数据(http://pds-geosciences.wustl.edu/missions/mro/spectral_library.htm[2017-03-20]),采用稀疏解混模型,对Eberswalde三角洲地区的矿物丰度进行定量反演,并根据矿物分布与含量分析当时可能的地质环境。
2 研究区与数据预处理
2.1 研究区介绍
Eberswalde撞击坑中心位于火星24.3°S,33.5°W,直径65.3 km,深度约800 m(图1(a))。地层、地貌、沉积物等证据显示在火星过去历史上该撞击坑长期存在液态水(Malin和Edgett,2003)。Eberswalde撞击坑是火星探测的热点地区,曾作为火星巡视器“好奇号”与“Mars 2020”的备选着陆区之一。故本文选取Eberswalde撞击坑西北部三角洲作为研究区,定量反演该地区的矿物含量。
2.2 实验数据
2.2.1 CRISM数据
CRISM搭载于2005年8月美国发射的火星侦察轨道器(MRO),其目标探测模式数据由可见近红外(VNIR:0.36—1.06 μm)和红外(IR:1.00—3.94 μm)两部分组成,共544个波段,光谱间隔6.55 nm,空间分辨率为18 m,覆盖火星表面2%左右的地区(Murchie 等,2007)。由于矿物的吸收特征大多位于短波红外谱段,且根据CRISM数据信噪比分析(Lin 等,2016),本文选取目标探测模式数据红外的6—236波段(1.034—2.549 μm)进行矿物丰度反演,使用CRISM Analysis Toolkit(CAT)完成格式转换、光度校正、大气校正、几何校正、去条带和光谱去噪等预处理流程,其中大气校正采用Volcano-Scan算法(McGuire 等,2009)。
2.2.2 CRISM光谱库数据
构建恰当合理的光谱库是稀疏解混算法的第一步,光谱库中的光谱要尽量和高光谱传感器的测量条件和环境一致,且需囊括目标地物中可能存在的所有矿物光谱。CRISM光谱库是CRISM科学团队提供的一套实验室模拟的火星样本(由混合物、矿物、有机物3部分组成)光谱数据集。目前包含1134个自然或合成的火星模拟样品的2260条光谱。本文选取库中纯净矿物光谱738条,其中包括硅酸盐、硫酸盐、氧化物、氢氧化物、碳酸盐、卤化物、磷酸盐、硝酸盐矿物。这些光谱在干燥条件下通过双锥反射测量获得,符合火星表面的干燥环境(Lin 等,2016)。采用线性插值将所有矿物光谱重采样至CRISM数据的中心波长。
3 原理与方法
稀疏分解直接利用光谱库内的纯净端元光谱来进行线性解混,但由于光谱库数量庞大、互信息度高(Iordache 等,2012a),且“相似”光谱端元会产生扰动,影响解的稳定性。为了降低运算时间,提高解混精度,本文通过高光谱数据子空间识别算法(Hysime)(Bioucas-Dias和Nascimento,2008)进行CRISM光谱库约简,然后利用协同稀疏分解(CLSUnSAL)(Iordache 等,2014a)进行研究区矿物定量反演。
3.1 光谱库约简方法
假定一幅高光谱数据Y可表示为Y=S+N,其中
${S} = {A} \times {X}$ | (1) |
式中,A是l×p满秩矩阵,p<1,在线性解混情况下,A是端元光谱;X是对应的丰度矩阵,X服从和为1和非负约束。
Hysime分别估计信号与噪声的相关矩阵,根据最小均方误差的原则选择信号子空间的最优特征向量子集(Bioucas-Dias和Nascimento,2008)。
本文使用高光谱信号子空间识别(Hysime)算法(Bioucas-Dias和Nascimento,2008)获取CRISM数据端元个数k和投影矩阵
${\varepsilon _j} = {\left\| {P}{A}_j \right\|_2}/{\left\| {A}_j \right\|_2}$ | (2) |
式中,
3.2 协同稀疏解混原理
协同稀疏分解(CLSUnSAL)又叫多任务(Multitask)稀疏分解,考虑到图像中实际存在的端元数量往往远小于光谱库中光谱的个数,对应的丰度矩阵中存在很少的非零行。原理如图2(Iordache 等,2014a),高光谱图像中存在的端元在光谱库中呈绿色,对应的丰度矩阵行呈蓝色。多任务稀疏改变了逐像元分解的模式,同时对所有像元的端元光谱数进行约束,公式为(Iordache 等,2014a)
$\mathop {\min }\limits_x {\left\| {AX} - {Y} \right\|_{{F}}} + \lambda {{\left\| {X} \right\|}_{{{2}},{1}}} + {{{\kappa _{{R}}}_+ }}\left( {X} \right)$ | (3) |
式中,
4 实验与结果分析
4.1 实验结果
本文基于Hysime算法,得出FRT000060DD数据信号子空间的维数为8,即端元数k为8。将CRISM光谱库投影到子空间上并对投影误差排序,投影误差越小表示与遥感图像的端元光谱越相似,越有可能存在于图像中。由于大气校正、去噪可能存在残差,实际保留的端元光谱数是根据图像的估计端元个数、光谱库的光谱数量、投影误差设定的一个经验值,一般略大于k远小于m。取值有两种方法:(1)设定阈值,保留光谱库中所有投影误差小于阈值的光谱(Themelis 等,2010);(2)设定百分比,取投影误差最小的前n个光谱(n=m×b,m是光谱库的光谱总数,b是设定的百分比)(Ertürk 等,2017)。因为对不同高光谱数据和光谱库,投影误差值变化较大,本文设定百分比,选取了投影误差最小的前4%,即30条光谱作为精简后的端元光谱库,利用协同稀疏分解算法进行丰度反演,合并同种矿物并剔除最大丰度低于0.01%的矿物(认为不存在),得到的矿物名称及丰度如表1。为突出矿物与地形的关系,将矿物丰度图层叠加在CRISM数据上,颜色显示方案R(矿物丰度图层),G(1.506 μm反射率),B(1.080 μm反射率),6种矿物的分布及丰度分别如图3和表1所示。
表 1 矿物名称及对应丰度
Table 1 Minerals and corresponding abundance
英文名称 | 中文名称 | 化学式 | 矿物类型 | 丰度范围/% | 平均丰度/% |
Plagioclase | 斜长石 | (Na2, Ca)O·Al2O3·6SiO2 | 架状硅酸盐 | 0—59.4 | 24.5 |
Olivine | 橄榄石 | (Mg, Fe)2SiO4 | 岛状硅酸盐 | 0—57.2 | 18.8 |
Siderite | 菱铁矿 | FeCO3 | 碳酸盐 | 0—63.3 | 1.6 |
Diaspore | 硬水铝石 | HAlO2 | 氧化物 | 0—73.2 | 10.9 |
Pyroxene | 辉石 | Ca(Mg, Fe, A1)(Si, A1)2O6 | 链状硅酸盐 | 0—79.3 | 27.7 |
Tremolite | 透闪石 | Ca2(Mg, Fe)5Si8O22(OH)2 | 链状硅酸盐 | 0—42.5 | 15.8 |
4.2 分析与验证
由丰度反演结果可知,(1)图中存在5种原生矿物:辉石、橄榄石、斜长石、菱铁矿、硬水铝石和1种蚀变矿物透闪石。(2)斜长石主要分布在西部三角洲地区以及东北角一小型撞击坑边缘;橄榄石分布在中央地区,呈东西分布;辉石和菱铁矿主要分布于冲积扇前缘;硬水铝石主要沿河谷分布,东北角相对富集的硬水铝石可能是过去液态水搬运所致;透闪石主要分布在东北角。Poulet等人(2014)基于Shkuratov模型,对Eberswalde三角洲121个像元的平均光谱进行拟合,反演得到4种矿物及对应丰度:斜长石(51%)、高钙辉石(29%)、低钙辉石(6%)、锂蒙脱石(10%)。本文反演结果中也探测到斜长石、辉石矿物。由于Poulet等人(2014)仅在出露点上进行分析,与本文研究尺度不一致,且锂蒙脱石丰度较低,本文尚未发现锂蒙脱石。
斜长石是基本的造岩矿物,相对稳定;橄榄石和辉石属于不稳定矿物,在搬运和沉积过程中会发生蚀变,所以橄榄石和辉石分布的区域靠近物源区,原始物源分别为超基性岩(如橄榄岩)和基性岩(如辉长岩);菱铁矿化学成分是碳酸亚铁,反映还原的碳酸盐环境,可能为氧气不足、碳酸盐较丰富的浅海或者盐湖;硬水铝石是长石等矿物的风化产物,推测存在铝土型风化,流体将原始矿物中的Na/K/Mg等离子带走,残留Al离子;透闪石可能形成于接触变质作用中,岩浆提供高温含Si质的流体,碳酸盐围岩提供Ca质。
光谱检验是火星矿物反演常用的验证方法。火星表面矿物可能被灰尘覆盖,导致矿物光谱特征不明显,可在图像中选择不含该种矿物且趋势平坦的光谱做比值,比值光谱起到去除背景突出矿物光谱特征的作用(Viviano 等,2013)。菱铁矿由于碳酸根的倍频与合频振动,在1.92 μm、2.35 μm具有吸收特征;硬水铝石是铝的氧化物矿物,由于羟基振动在1.4 μm,1.81 μm具有吸收谱带(燕守勋 等,2003)。本文从图像中分别提取菱铁矿与硬水铝石丰度最高位置的光谱S1与丰度为0的光谱S2(X,Y是所提取光谱在图像上的行列坐标),并将比值光谱S1/S2与标准库光谱进行对比。菱铁矿(图4(a))的比值光谱和光谱库曲线趋势和吸收位置一致,而硬水铝石(图4(a))两者尽管趋势一致但比值光谱吸收位置不明显,可能由于火星矿物被表面灰尘覆盖所致。
CRISM光谱指数是进行火星表面矿物分布快速定性评估的有效手段,一般光谱指数值越大表明该种矿物含量越高。橄榄石指数、斜长石指数、辉石指数分别探测中心波长位于1 μm(式4)、1.3 μm(式5)、1.81 μm(式6)的吸收特征,式中RB=1–CR,CR为反射光谱去包络线后数值(Viviano-Beck 等,2014)。对比本文反演的3种主要矿物斜长石、辉石与橄榄石的丰度图与光谱指数图(图5)。本文解混结果与光谱指数图分布趋势一致,且散点图显示每种矿物光谱指数与丰度呈线性关系。
$\begin{array}{l}(R{B_{1080}} + R{B_{1152}} + R{B_{1210}} + R{B_{1250}}) \cdot 0.03 + \\[8pt](R{B_{1263}} + R{B_{1276}}) \cdot 0.07 + (R{B_{1330}} + R{B_{1368}}) \cdot 0.12 + \\[8pt]R{B_{1395}} \cdot 0.14 + (R{B_{1427}} + R{B_{1470}}) \cdot 0.18\end{array}$ | (4) |
$1 - {R_{1320}}/\left( {0.64 {R_{1080}} + 0.36 {R_{1750}}} \right)$ | (5) |
$\left( {R{B_{1690}} + R{B_{1750}}} \right) \cdot 0.2 + \left( {R{B_{1810}} + R{B_{1870}}} \right) \cdot 0.3$ | (6) |
5 结 论
火星表面矿物研究对于了解火星的地质演化和圈定过去液态环境范围具有重要意义。本文基于CRISM高光谱数据和CRISM光谱库,先采用光谱库约简再使用协同稀疏算法对Eberswalde撞击坑三角洲的矿物丰度进行定量反演,得到辉石、橄榄石、斜长石、菱铁矿、硬水铝石和透闪石6种矿物。其中橄榄石和辉石不稳定,指示其分布的区域靠近物源,原始物源分别为超基性岩(如橄榄岩)和基性岩(如辉长岩);冲积扇地区存在的矿物在冲积扇外也有分布,可能是流体搬运所致;该区域的透闪石可能形成于接触变质作用中,岩浆提供高温含Si质的流体,碳酸盐围岩提供Ca质。该地区已有矿物丰度反演研究仅对三角洲出露点121个像元的平均光谱进行反演,与本文研究尺度不一致,且锂蒙脱石丰度较低,本文尚未发现锂蒙脱石。
随着火星探测不断深入,构建更加完备的光谱库,涵盖火星表面更多地物的光谱信息成为可能;火星和地球大气环境存在差异,发展新的大气校正算法将有望从CRISM数据获取更加精确的光谱;此外,考虑入射光线在不同矿物组分间多次反射引起的非线性混合影响,这些都会提高火星表面矿物丰度反演精度,也会进一步促进对火星地质环境的认识。
参考文献(References)
-
Bandfield J L. 2002. Global mineral distributions on Mars. Journal of Geophysical Research, 107 (E6): 9–1. [DOI: 10.1029/2001je001510]
-
Bibring J P, Langevin Y, Mustard J F, Poulet F, Arvidson R, Gendrin A, Gondet B, Mangold N, Pinet P, Forget F and The OMEGA Team. 2006. Global mineralogical and aqueous mars history derived from OMEGA/Mars Express data. Science, 312 (5772): 400–404. [DOI: 10.1126/science.1122659]
-
Bioucas-Dias J M and Figueiredo M A T. 2010. Alternating direction algorithms for constrained sparse regression: application to hyperspectral unmixing//Proceedings of the 2nd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing. Reykjavik, Iceland: IEEE: 1–4 [DOI: 10.1109/WHISPERS.2010.5594963]
-
Bioucas-Dias J M and Nascimento J M P. 2008. Hyperspectral subspace identification. IEEE Transactions on Geoscience and Remote Sensing, 46 (8): 2435–2445. [DOI: 10.1109/TGRS.2008.918089]
-
Bishop J L, Loizeau D, McKeown N K, Saper L, Dyar M D, Des Marais D J, Parente M and Murchie S L. 2013. What the ancient phyllosilicates at Mawrth Vallis can tell us about possible habitability on early Mars. Planetary and Space Science, 86 : 130–149. [DOI: 10.1016/j.pss.2013.05.006]
-
Carter J, Poulet F, Murchie S and Bibring J P. 2013. Automated processing of planetary hyperspectral datasets for the extraction of weak mineral signatures and applications to CRISM observations of hydrated silicates on Mars. Planetary and Space Science, 76 : 53–67. [DOI: 10.1016/j.pss.2012.11.007]
-
Combe J P, Mouélic S L, Sotin C, Gendrin A, Mustard J F, Deit L L, Launeau P, Bibring J P, Gondet B, Langevin Y, Pinet P and The OMEGA Science Team. 2008. Analysis of OMEGA/Mars Express data hyperspectral data using a Multiple-Endmember linear spectral unmixing model (MELSUM): methodology and first results. Planetary and Space Science, 56 (7): 951–975. [DOI: 10.1016/j.pss.2007.12.007]
-
Ertürk A, Iordache M D and Plaza A. 2017. Sparse unmixing with dictionary pruning for hyperspectral change detection. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 10 (1): 321–330. [DOI: 10.1109/JSTARS.2016.2606514]
-
Iordache M D, Bioucas-Dias J M and Plaza A. 2011. Sparse unmixing of hyperspectral data. IEEE Transactions on Geoscience and Remote Sensing, 49 (6): 2014–2039. [DOI: 10.1109/TGRS.2010.2098413]
-
Iordache M D, Bioucas-Dias J M and Plaza A. 2012a. Dictionary pruning in sparse unmixing of hyperspectral data. Proceedings of the 4th Workshop on Hyperspectral Image and Signal Processing. Shanghai, China: IEEE: 1–4 [DOI: 10.1109/WHISPERS.2012.6874329]
-
Iordache M D, Bioucas-Dias J M and Plaza A. 2012b. Total variation spatial regularization for sparse hyperspectral unmixing. IEEE Transactions on Geoscience and Remote Sensing, 50 (11): 4484–4502. [DOI: 10.1109/TGRS.2012.2191590]
-
Iordache M D, Bioucas-Dias J M and Plaza A. 2014a. Collaborative sparse regression for hyperspectral unmixing. IEEE Transactions on Geoscience and Remote Sensing, 52 (1): 341–354. [DOI: 10.1109/TGRS.2013.2240001]
-
Iordache M D, Bioucas-Dias J M, Plaza A and Somers B. 2014b. MUSIC-CSR: hyperspectral unmixing via multiple signal classification and collaborative sparse regression. IEEE Transactions on Geoscience and Remote Sensing, 52 (7): 4364–4382. [DOI: 10.1109/TGRS.2013.2281589]
-
Lewis K W and Aharonson O. 2006. Stratigraphic analysis of the distributary fan in Eberswalde crater using stereo imagery. Journal of Geophysical Research, 111 (E6): E06001 [DOI: 10.1029/2005JE002558]
-
Lin H L and Zhang X. 2017. Retrieving the hydrous minerals on Mars by sparse unmixing and the Hapke model using MRO/CRISM data. Icarus, 288 : 160–171. [DOI: 10.1016/j.icarus.2017.01.019]
-
Lin H L, Zhang X and Sun Y L. 2016. Hyperspectral sparse unmixing of minerals with single scattering albedo. Journal of Remote Sensing, 20 (1): 53–61. [DOI: 10.11834/jrs.20165079] ( 林红磊, 张霞, 孙艳丽. 2016. 基于单次散射反照率的矿物高光谱稀疏解混. 遥感学报, 20 (1): 53–61. [DOI: 10.11834/jrs.20165079] )
-
Lin H L, Zhang X, Shuai T, Zhang L F and Sun Y L. 2016. Abundance retrieval of hydrous minerals around the Mars Science Laboratory landing site in Gale crater, Mars. Planetary and Space Science, 121 : 76–82. [DOI: 10.1016/j.pss.2015.12.007]
-
Malin M C and Edgett K S. 2003. Evidence for persistent flow and aqueous sedimentation on early Mars. Science, 302 (5652): 1931–1934. [DOI: 10.1126/science.1090544]
-
McGuire P C, Bishop J L, Brown A J, Fraeman A A, Marzo G A, Frank Morgan M, Murchie S L, Mustard J F, Parente M, Pelkey S M, Roush T L, Seelos F P, Smith M D, Wendt L and Wolff M J. 2009. An improvement to the volcano-scan algorithm for atmospheric correction of CRISM and OMEGA spectral data. Planetary and Space Science, 57 (7): 809–815. [DOI: 10.1016/j.pss.2009.03.007]
-
Moussaoui S, Hauksdóttir H, Schmidt F, Jutten C, Chanussot J, Brie D, Douté S and Benediktsson J A. 2008. On the decomposition of Mars hyperspectral data by ICA and Bayesian positive source separation. Neurocomputing, 71 (10-12): 2194–2208. [DOI: 10.1016/j.neucom.2007.07.034]
-
Murchie S, Arvidson R, Bedini P, Beisser K, Bibring J P, Bishop J, Boldt J, Cavender P, Choo T, Clancy R T, Darlington E H, Des Marais D, Espiritu R, Fort D, Green R, Guinness E, Hayes J, Hash C, Heffernan K, Hemmler J, Heyler G, Humm D, Hutcheson J, Izenberg N, Lee R, Lees J, Lohr D, Malaret E, Martin T, McGovern J A, McGuire P, Morris R, Mustard J, Pelkey S, Rhodes E, Robinson M, Roush T, Schaefer E, Seagrave G, Seelos F, Silverglate P, Slavney S, Smith M, Shyong W J, Strohbehn K, Taylor H, Thompson P, Tossman B, Wirzburger M and Wolff M. 2007. Compact reconnaissance imaging spectrometer for mars (CRISM) on mars reconnaissance orbiter (MRO). Journal of Geophysical Research, 112 (E5): E05S03 [DOI: 10.1029/2006JE002682]
-
Mustard J F, Ehlmann B L, Murchie S L, Poulet F, Mangold N, Head J W, Bibring J P and Roach L H. 2009. Composition, morphology, and stratigraphy of Noachian Crust around the Isidis basin. Journal of Geophysical Research, 114 (E2): E00D12 [DOI: 10.1029/2009JE003349]
-
Ody A, Poulet F, Langevin Y, Bibring J P, Bellucci G, Altieri F, Gondet B, Vincendon M, Carter J and Manaud N. 2012. Global maps of anhydrous minerals at the surface of Mars from OMEGA/MEx. Journal of Geophysical Research, 117 (E11): E00J14 [DOI: 10.1029/2012JE004117]
-
Pelkey S M, Mustard J F, Murchie S, Clancy R T, Wolff M, Smith M, Milliken R, Bibring J P, Gendrin A, Poulet F, Langevin Y and Gondet B. 2007. CRISM multispectral summary products: parameterizing mineral diversity on Mars from reflectance. Journal of Geophysical Research, 112 (E8): E08S14 [DOI: 10.1029/2006je002831]
-
Poulet F, Carter J, Bishop J L, Loizeau D and Murchie S M. 2014. Mineral abundances at the final four curiosity study sites and implications for their formation. Icarus, 231 : 65–76. [DOI: 10.1016/j.icarus.2013.11.023]
-
Schmidt F, Bourguignon S, Le Mouëlic S, Dobigeon N, Theys C and Tréguier E. 2011. Accuracy and performance of linear unmixing techniques for detecting minerals on OMEGA/Mars express//Proceedings of the 3rd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing. Lisbon, Portugal: IEEE: 1–4 [DOI: 10.1109/WHISPERS.2011.6080858]
-
Tang W, Shi Z W, Wu Y and Zhang C S. 2015. Sparse unmixing of hyperspectral data using spectral a priori information. IEEE Transactions on Geoscience and Remote Sensing, 53 (2): 770–783. [DOI: 10.1109/TGRS.2014.2328336]
-
Themelis K, Rontogiannis A A, Koutroumbars K. 2010. Semi-Supervised hyperspectral unmixing via the weighted lasso. IEEE International Conference on Acoustics, Speech and Signal Processing, Texas: 1194–1197
-
Viviano C E, Moersch J E and McSween H Y. 2013. Implications for early hydrothermal environments on Mars through the spectral evidence for carbonation and chloritization reactions in the Nili Fossae region. Journal of Geophysical Research, 118 (9): 1858–1872. [DOI: 10.1002/jgre.20141]
-
Viviano-Beck C E, Seelos F P, Murchie S L, Kahn E G, Seelos K D, Taylor H W, Taylor K, Ehlmann B L, Wiseman S M, Mustard J F and Morgan M F. 2014. Revised CRISM spectral parameters and summary products based on the currently detected mineral diversity on Mars. Journal of Geophysical Research, 119 (6): 1403–1431. [DOI: 10.1002/2014je004627]
-
Yan S X, Zhang B, Zhao Y C, Zheng L F, Tong Q X and Yang K. 2003. Summarizing the VIS-NIR spectra of minerals and rocks. Remote Sensing Technology and Application, 18 (4): 191–201. [DOI: 10.3969/j.issn.1004-0323.2003.04.002] ( 燕守勋, 张兵, 赵永超, 郑兰芬, 童庆禧, 杨凯. 2003. 矿物与岩石的可见—近红外光谱特性综述. 遥感技术与应用, 18 (4): 191–201. [DOI: 10.3969/j.issn.1004-0323.2003.04.002] )