【作者】 田丰；

【导师】 王润生；

【作者基本信息】 中国地质大学（北京） ， 地图制图学与地理信息工程， 2010， 博士

【摘要】 高光谱遥感矿物填图发展到现在,主要研究集中在可见光—反射红外(0.35～2.5μm)波段,矿物种类填图技术方法已逐渐成熟,形成了一套较完善的方法体系。矿物含量定量反演也取得了一定的进展,但也面临着许多新的挑战,如(1)对岩石和矿物介质表面与光辐射之间相互作用的机理研究还比较滞后,而对地物介质结构信息的光学特性时空分布规律和光场二向性分布规律的研究,能够提高遥感岩矿信息精细定量提取的精度,具有十分重要的科学和现实意义(。2)将高光谱矿物填图的光谱区间和提取技术从可见-反射红外波段延伸到中红外(2.5～25μm)波段,建立全谱段光谱矿物识别规则及方法可以全面提高遥感岩矿识别的能力与精度。但是,由于中热红外遥感数据获取、处理及发射率反演等问题难度较大,中红外波段的岩矿信息提取研究滞后。本次研究以多类型的矿物粉末为研究对象,在可见光—反射红外(0.35～2.5μm)波段,以实验室测量的多角度岩矿二向性反射率为数据源,使用Matlab和Visual c#语言开发了基于“有约束的非线性最小二乘法”的反演系统,提出并建立了基于“先验知识”基础上的参数分阶段反演和敏感性分析评价体系。将专家的先验知识、Hapke反射模型以及实验测量数据结合在一起。实现了对Hapke反射模型的五个参数(平均单次散射反照率W、相函数前后散射比例系数c、相函数振幅b、后向散射的经验系数B0、后向效应宽度h)在可见光—反射红外全波段内的初始值的敏感性分析和最优化反演。敏感性分析表明:平均单次散射反照率W、相函数前后散射比例系数c和相函数振幅b的反演受参数初始值影响较大,敏感性强;向散射的经验系数B0和后向效应宽度h的反演受参数初始值影响较小,敏感性较弱。确定参数的最优化反演初始值后的反演结果表明:反演结果均方根误差很小,Hapke反射模型模拟计算二向性反射率和原始数据误差很小,Hapke反射模型可以从物理层面上模拟岩矿反射光谱的形成及变异机理。总结了五个参数的反演值在在以岩矿介质为目标时的规律和特点,深刻理解这些参数对研究岩矿光谱机理和进一步提高和改进遥感岩矿定量识别精度有着重要意义。本次研究系统分析和总结了主要矿物在中热红外(2.5～25μm)区间光谱红外光谱机理和分类,对常见矿物中红外漫反射光谱进行了分析,表明在此区间可识别硅酸盐(包括不含水造岩矿物)、硫酸盐探测、碳酸盐、磷酸盐、氧化物、氢氧化物等矿物,遥感岩矿识别的能力和精度可以得到全面的提高。以实验室测量的多类型中热红外漫反射光谱为端元光谱和混合光谱,以基于光谱相似性算法的识别方法作为技术支撑,较深入地研究和开发了中热红外(2.5～25μm)区间矿物光谱混合特性、光谱解混及矿物含量反演的方法。综合上述研究表明,对中热红外区间的矿物进行种类识别和光谱解混及含量反演的研究和利用Hapke反射模型对岩矿的二向反射特性进行研究,是对现有的高光谱遥感矿物填图技术在深度和广度上的有益探索,在此基础上综合应用可见光-反射红外遥感和中热红外遥感开发全谱段矿物识别规则,进行岩矿识别是提高遥感岩矿识别精度、能力及可靠性的有效途径之一。更多还原

【Abstract】 By this time, the technology of hyperspectral mineral mapping mainly research on visible - infrared (0.35 ~ 2.5μm). The methods of the mineral mapping have gradually matured, and have formed a more complete methodology. Quantitative inversion of mineral content has also made progress, but it is also facing many new challenges, such as: (1) It relativly lag behind for the mechanization research between the rock and mineral medium surface and light radiation. But it is the basis of precise quantitative remote sensing information extraction of rock and mineral to the capture of the structural information and accurate bidirectional reflectance data. (2)Make the spectral region of hyperspectral mineral mapping and the extraction technology range from the visible - infrared reflection to extend to the mid-infrared (2.5 ~ 25μm) band, set up the whole spectrum mineral identification rules and method, in order to improve the abillity and precision of rock recognition base on remote sensing. However, duing to the thermal infrared remote sensing data acquisition, processing and emissivity data inversion is a great problem, which lead to the research of the information extraction in the middle-infrared band lag.Many types of mineral powder is regarded as the research object. In visible -reflection infrared (0.35～2.5μm) band, multi-angle rock bidirectional reflectance measured in laboratory was the data source. This paper developed the inversion system based on "a constrained nonlinear least squares method" using Matlab and Visual c # language, proposed and established sensitivity analysis and retrieval system in stages based on "prior knowledge". The expert’s prior knowledge, Hapke reflectance model and the experimental data were hung together, which realized the sensitivity analysis and optimization inversion of the initial value in all visible light - reflection Infrared(0.35 ~ 2.5μm) bands for the five parameters of Hapke reflectance model (the average single scattering albedo W, before and after the scattering phase function scale factor c, the amplitude of the phase function b, the backscatter experience coefficient B0, the width of the effect after h) .Sensitivity analysis shows that: The contribution of the various parameters of Hapke reflectance model to the bidirectional reflectance is interactive. The multiple parameters involved in inversion is an interaction. The relevance of the sensitivity of inversion parameters is also an important influence factor for parameter inversion accuracy. The average single scattering albedo W, before and after the scattering phase function scale factor , the amplitude of the phase function b have high sensitivity on the set of initial value, the backscatter experience coefficient B0, and the width of the effect after h has less sensitivity on the set of the initial value.The inversion results of the optimum inversion initial value after determining parameters show that: the root mean square error of the inversion results is very small. The bidirectional Reflectance simulated by Hapke reflectance model is almost the same as the raw data. This explained that Hapke reflectance model can simulate the spectrum formation and the variation mechanisms from the physical level. This paper summarized the laws and characteristics of the five parameters based on the target of rock. It plays an important role on understanding the rock and mineral spectral mechanism and the improvement of the quantitative identification about the rock and minerals based on remote sensing.Analyzing and summarizing the IR spectral mechanism and classification of the major minerals in the thermal infrared (2.5 ~ 25μm). The infrared spectra analysis of common minerals showed that silicate, sulfate detection, carbonates, phosphates, oxides, hydroxides and other minerals can be identified in this interval (including non-water rock-forming minerals),. The identification capabilities and precision of remote sensing of rock and minerals can be comprehensively improved. Taking the multi-types middle-thermal infrared spectra measured by the lib as the endmember spectra and mixed spectra, taking identification methods based on spectral similarity algorithm as the technical support, deeply researched and developed the mineral spectral mixed properties, mineral unmixing and the inversion method of mineral content in the thermal infrared (2.5 ~ 25μm ).The above study shows that, it is a useful exploration in the depth and breadth for the existing hyperspectral remote sensing mineral mapping to study the mineral species identification, spectral unmixing, content inversion on the range of thermal infrared and study the bidirectional reflectance characteristics of rock and mineral using Hapke reflectance model. Based on these, comprehensively applying the visible t - reflection infrared remote sensing and thermal infrared spectral to develop the whole spectrum mineral identification rules. Rock and mineral identification is an effective way to improve the accuracy, capability and reliability of rock and mineral identification of remote sensing.更多还原