摘要:
月表主要矿物含量的定量反演是月球科学领域的重要课题之一,对未来月表矿物信息解译有重要指导意义,因此提出一种对月表高光谱遥感数据线性分解获取矿物含量的方法.首先,利用Hapke辐射传输模型将Relab光谱库中的5种矿物(单斜辉石、斜方辉石、斜长石、橄榄石和钛铁矿)非线性混合的反射光谱转换为线性混合的单次反照率;然后,按照比例随机生成混合像元;最后基于全约束线性光谱分解方法建立上述5种矿物分解含量与真实含量的统计关系模型.利用Apollo登陆采样点实测数据对该模型进行验证的结果表明,辉石、斜长石、橄榄石和钛铁矿的反演结果与实测结果的相关系数分别为0.83,0.86,0.72和0.77.采用上述方法,利用印度探月卫星Chandrayaan-1搭载的月球矿物制图仪(moon mineralogy mapper,M3)高光谱数据得到月表虹湾地区矿物的含量分布图,表明利用全约束线性分解对高光谱矿物识别和含量反演是一种行之有效的方法.
Abstract:
Quantitative inversion of the main lunar mineral abundance is an important issue in the field of lunar science,which has great guiding effect on lunar mineral information interpretation in the future.In this paper,a method of linear unmixing for hyperspectral remote sensing image is proposed to obtain the mineral abundance.Firstly,based on Hapke radiative transfer model,the authors transformed the non-linear mixed reflectance spectra into the linear mixed single scattering albedo of five end-member minerals (clinopyroxene,orthopyroxene,plagioclase,olivine and ilmenite) in the Relab spectral library,and then generated mixed pixels according to the linear proportion randomly;in this way,the statistical relationship models of mineral unmixing abundance and real abundance for the above five minerals can be established respectively based on fully constrained linear-unmixing method.The verification result with the data measured by Apollo sampling points indicates that the correlation coefficient between mineral abundances inversed and measured in Apollo of pyroxene,plagioclase,olivine and ilmenite are 0.83,0.86,0.72 and 0.77 respectively.With this method,the authors acquired the lunar mineral abundance distribution maps in Sinus Iridum using Chandrayaan-1 M3 hyperspectral data,which shows that the fully constrained linear-unmixing method is a kind of effective approach to mineral identification and inversion with hyperspectral data.
下载: