Accuracy analysis of lunar mineral endmembers extraction using simulated Chang＇ E-1 IIM data  

作　　者：帅通[1,2] 张霞[1] 张明[1,2] 刘波[3] BAIG Muhammadhasanali[1,2,4] 

机构地区：[1]遥感科学国家重点实验室中国科学院遥感应用研究所,北京100101 [2]中国科学院研究生院,北京100039 [3]环境保护部南京环境科学研究所,南京210042 [4]巴基斯坦气象局

出　　处：《遥感学报》2012年第6期1205-1221,共17页NATIONAL REMOTE SENSING BULLETIN

基　　金：国家高技术研究发展计划(863计划)(编号:2009AA122203);遥感科学国家重点实验室自由探索基金(编号:10ZY-04)~~

摘　　要：干涉型成像光谱仪(IIM)作为中国首颗探月卫星嫦娥一号上的重要载荷之一,已经获取了大量的月球高光谱数据。然而由于IIM数据的光谱范围被限定在480—960nm波谱范围内,很少覆盖月表矿物的波谱吸收特征,利用IIM数据反演月表矿物信息面临巨大挑战。当前月表地面调查数据或精细矿物分布制图还很缺乏,难以对反演或提取结果进行有效验证。本文利用月表4种主要矿物(斜长石、单斜辉石、橄榄石和钛铁矿)进行线性混合构建IIM模拟数据,分析月表矿物端元提取的精度,并对真实IIM数据进行了初步应用。根据斜长石和钛铁矿的不同组合比例分别模拟了月表的月海和高地,模拟数据还加入了与IIM真实数据分布趋势一致的随机噪声。基于4种端元提取算法(VCA、ICA、MVSA、SISAL)对模拟数据进行端元提取,并以光谱角距离(SAD)作为端元提取的精度评价准则。结果显示,MVSA和SISAL算法提取的所有矿物端元的光谱角距离都小于0.1,一般意味着矿物提取精度可以被接受。斜长石在各种算法下提取端元的光谱角距离都小于0.015;单斜辉石和钛铁矿提取端元的光谱角距离都小于0.1;橄榄石提取精度最差,在ICA算法下其光谱角距离高达0.34。As one of the most important payloads on Chang＇E-1 satellite, interference imaging spectrometer （IIM） has acquired large amount of hyperspectral data. However, it is still a great challenge to retrieve minerals distribution information of the lu- nar surface due to the limited spectral wavelength range of IIM data （480-960 nm） which does not cover spectral absorption features of lunar minerals completely. At present, precise quantitative mineral distributions or credible surveys of lunar surface are lacking to validate the retrieved results. In this paper, we perfomed simulation of IIM spectral data based on spectral linear mixing method with four main lunar minerals （plagioclase, clinopyroxene, olivine and ilmenite）. We analyzed the accuracy of mineral endmembers extraction and implementeda preliminary application with true IIM data. We simulated the spectra of lunar mare and highland according to different proportion of plagioclase and ilmenite respectively. We added random noises of the same trend as IIM and employed four endmember extraction approaches （vertex component analysis, independent component analysis, minimum volume simplex analysis, simplex identification via split augmented lagrangian to extract endmembers. We used spectral angle distance （SAD） as a criteria for accuracy analysis. The results show that all the SADs of endmembers ex- tracted by MVSA and SISAL are lower than 0.1. Plagioclase （the SADs that are lower than 0.015） can always be extracted by all approaches. All SADs of clinopyroxene and ilmenite are lower than 0. l with olivine having the lowest extraction accuracy.

关 键 词：遥感技术 遥感方式 遥感图像 应用 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]