植被覆盖区矿产勘查中高光谱遥感蚀变信息提取研究

In application of hyperspectral remote sensing (RS) for mineral exploration in vegetation covered area, vegetation will weaken information of alteration minerals in pixel spectrum so that the information is difficult to be effectively extracted from the image. In addition to that, the extracted information of alteration minerals contains much useless information for prospecting prediction. In order to successfully extract useful information for prospecting prediction, this research will select the Pulang copper deposit and its surrounding area of Yunan province as the study area, and identify altered rocks which are related to the genesis of mineral deposit of the study area. This research will firstly collect field and laboratory spectra of different types of altered rocks under different proportions of vegetation; then using these spectra and the hyperspectral RS data of the study area, the steady bands’ features which mean types of altered rocks will be obtained by analysing these spectral curves or their transform curves, these features’ parameters will be quantified, and the inevitability of the features existed in the mixed spectra will be explained from the perspective of physics. Using these features’ parameters, the alteration information will be extracted from the hyperspectral image, and it will be compared with the information by the conventional methods. In this study, important task is to obtain these steady bands’ features and their quantified parameters. The results of this study will be useful for accurate identification of rock and mineral in hyperspectral imagery, and for effective delineation of useful regions for prospecting prediction in vegetation covered area.

高光谱遥感在植被覆盖区矿产勘查应用中，植被等会造成像元光谱中蚀变矿物信息微弱，不易被有效提取，且提取的蚀变矿物信息包含了很多对找矿无用的信息。为了成功提取与找矿相关的蚀变信息，该研究以云南普朗铜矿区及其外围地区为实验区，进行与实验区成矿相关的蚀变岩信息提取。首先进行任意程度植被覆盖下不同类型蚀变岩的野外和实验室光谱测试，然后综合利用这些测试光谱和高光谱遥感影像数据，探讨确定混合光谱曲线或其变换曲线中对蚀变岩有指示意义的“稳定谱段特征”，并对特征参数进行定量描述，对稳定谱段特征存在的必然性进行物理解释。基于这些特征参数，进行高光谱遥感找矿蚀变信息提取，并与常规方法的结果进行对比分析。该研究中，混合光谱中稳定谱段的确定及特征描述是重点研究内容。该研究对于提高植被覆盖区岩矿光谱的识别效果和找矿有利地带圈定的可靠性具有现实意义。

高光谱遥感在植被覆盖区矿产勘查应用中，植被等会造成像元光谱中蚀变矿物信息微弱，不易被有效提取。为了成功提取与找矿相关的蚀变信息，该研究以云南普朗铜矿区及其外围地区为实验区，进行与实验区成矿相关的蚀变岩信息提取。主要研究内容包括：蚀变岩及植物的光谱测试、混合光谱的稳定谱段的确定及特征描述、稳定谱段及特征的物理解释、基于稳定谱段特征的高光谱遥感影像蚀变信息提取、信息提取结果检验。重要结果有：（1）傅里叶变换曲线0hz处的振幅远大于1hz之后（含1hz）的振幅，50hz之后各蚀变类型的振幅都很小且变化较小，可选用变换曲线1hz-50hz的部分进行特征描述；（2）相同原岩不同蚀变的光谱的变换曲线形态比较接近，不同原岩光谱的变换曲线的形态差异明显，各种蚀变类型的变换曲线在前30hz范围内差异较大，而在接近50hz时趋于相同；（3）混合比（植被/岩石-植被）在40%-60%时，线性光谱模型模拟精度低于其他混合比，稳定谱段特征的“稳定”程度不如其它混合比；（4）实现了基于稳定谱段的单一特征识别方法、多种特征相结合识别方法，建立了含硅化、黄铁矿化、黄铜矿化和绢云母化的青磐岩化石英闪长玢岩及石英二长斑岩光谱识别模型；（5）将基于Symlet小波变换的ResNet模型应用于蚀变岩光谱的识别，研究表明该模型分类准确率均高于其它模型；（6）进行基于空谱残差网络的高光谱影像岩性分类与识别研究，研究结果明显好于1D-CNN模型和2D-CNN模型的结果，这表明空谱残差网络模型更适用于高光谱影像岩性分类。该研究对于提高植被覆盖区岩矿光谱的识别效果和找矿有利地带圈定的可靠性具有现实意义。