KMS XINJIANG INSTITUTE OF ECOLOGY AND GEOGRAPHY，CAS
| 基于地面实测光谱的矿区 Cu 元素定量反演研究 | |
| Alternative Title | Quantitative Inversion of Cu Element in Mining Area Based on Ground Measured Spectrum |
| 马秀梅 | |
| Subtype | 硕士 |
| Thesis Advisor | 周可法 |
| 2019-06-30 | |
| Degree Grantor | 中国科学院大学 |
| Place of Conferral | 北京 |
| Degree Discipline | 工学硕士 |
| Keyword | 地面实测光谱 岩石 Cu 元素 PLSR 定量反演 Ground Measured Spectrum Rock Cu Element PLSR Quantitative Inversion |
| Abstract | 矿产资源对一个国家的重要性日渐凸显出来， 如何快速、大面积获取地球化学元素信息成为关键， 因此将遥感与地球化学相结合成为必然的发展趋势。本文以卡拉塔格玉带斑岩铜矿床、卡拉麦里矿集区为研究区， 以探寻偏最小二乘法（PLSR） 最优建模参数为目标，利用 PLSR 建模，针对矿区周围不同风化程度的岩石，基于地面实测光谱建立 Cu 元素的定量反演模型，生成实测值与预测值的含量分布图，对比高低值区域是否吻合。岩石中金属元素含量一般较低，很难形成对应的特征波段，但是可以寻找与金属元素最相关的波段。本文主要针对岩石中的 Cu 元素，首先探讨了关于岩石中金属元素与 Cu 元素的相关关系，并找出了与 Cu 元素具有极显著相关关系的元素有 Cr、 Fe、 Zn、 V，并提取出了与 Cu 元素相关关系最高的波段有 Cu 相关系数较高的为 600~750 nm、 1000 nm、 1900~2100 nm、 2250 nm、 2350 nm 等波段范围的附近，并找出了其余四个与 Cu 元素具极显著相关关系的元素的相关系数较高的波段，这为以后利用间接方法反演岩石中 Cu 元素打下基础。通过建立岩石中 Cu 元素的 PLSR 定量反演模型，探讨了不同波段宽度下PLSR 模型的精度，并得到在波段宽度在 30 nm 时，模型的精度最高且不会过拟合； 而对于岩石中 Cu 元素的最优波段选取，通过逐步回归法，入选的波段依次为： 670 nm、 679 nm、 749 nm、 751 nm、 752 nm、 758 nm、 770 nm、 1946 nm、2127 nm、 2136 nm。 建立 PLSR 模型时，研究出当波段数为 9 个时，模型的精度最优且不会出现过拟合。对于不同的条件下，岩石中 Cu 元素反演的精度也不同。本文探讨了不同风化程度及不同岩性条件下，对建模精度的影响，发现除去结构对岩石的影响因素，岩石的新鲜面比风化面反演精度较高； 对于岩性对模型的影响，发现相比较所有岩性建模精度，利用硅质岩进行定量反演预测铜元素含量建模精度不高。将两种选取的不同最优参数建模定量反演 Cu 元素含量，并利用反距离权重法插值生成实测值的含量分布图与预测值的含量分布图，并交叉研究区验证两种不同参数的建模，对比实测值的含量分布图，得出结论，两种不同参数的建模，预测值与实测值的高低值区域大致相同，但相比较而言最优波段建模的预测值的高低值范围较贴近实测值的范围。为以后应用于高光谱遥感大面积反演 Cu 元素含量，并圈定异常区域提供了地面试验依据。 |
| Other Abstract | The importance of mineral resources to a country is increasingly prominent. Howto acquire information on geochemical elements in a fast and large area is the key.Therefore, combining remote sensing with geochemistry has become an inevitabledevelopment trend. In this paper, the Karatag jade belt porphyry copper deposit andthe Kalamaili ore concentration area are used as research areas to explore the optimalmodeling parameters of partial least squares (PLSR), and the PLSR model is used totarget different weathering degrees around the mining area. The rock is based on theground measured spectrum to establish a quantitative inversion model of Cu elements,and the content distribution map of measured and predicted values is generated tocompare whether the high and low values are consistent.The content of metal elements in rocks is generally low, and it is difficult to formcorresponding characteristic bands, but it is possible to find the bands most relevant tometal elements. In this paper, we focus on the Cu element in rock. Firstly, we discussthe relationship between metal elements and Cu elements in rocks, and find out thatthe elements with extremely significant correlation with Cu elements are Cr, Fe, Zn, V,and extract them. The band with the highest correlation with Cu has a high Cucorrelation coefficient of 600~750 nm, 1000 nm, 1900~2100 nm, 2250 nm, 2350 nm,and finds that the other four are significantly related to Cu. The correlation coefficientof the elements of the relationship is higher in the band, which lays the foundation forthe inversion of the Cu element in the rock by indirect methods.By establishing a PLSR quantitative inversion model of Cu in rock, the accuracyof the PLSR model under different band widths is discussed, and the model has thehighest accuracy and no over-fitting when the band width is 30 nm. For the Cuelement in the rock The optimal band selection, by stepwise regression method, theselected bands are: 670 nm, 679 nm, 749 nm, 751 nm, 752 nm, 758 nm, 770 nm, 1946nm, 2127 nm, 2136 nm. When establishing the PLSR model, it is studied that whenthe number of bands is nine, the accuracy of the model is optimal and there is noover-fitting.For different conditions, the accuracy of Cu element inversion in rock is alsodifferent. In this paper, the influence of different weathering degrees and different lithology conditions on the modeling accuracy is discussed. It is found that theinfluence of the structure on the rock is removed. The fresh surface of the rock hashigher precision than the weathering surface. For the influence of lithology on themodel, It is found that compared with all lithology modeling accuracy, thequantitative prediction of copper element content by siliceous rock is not accurate.The two selected optimal parameters were modeled to quantitatively invert theCu element content, and the inverse distance weight method was used to interpolate togenerate the content distribution map of the measured values and the contentdistribution map of the predicted values, and cross-research area to verify the twodifferent parameters. Modeling, comparing the content distribution map of themeasured values, it is concluded that the modeling of the two different parameters isroughly the same as the high and low values of the measured values, but the predictedvalues of the optimal band modeling are compared. The range of values is closer tothe range of measured values. In order to apply the large-area inversion of Cu elementcontent in high-spectral remote sensing and to define the anomalous area, the groundtest basis is provided. |
| Subject Area | 地球探测与信息技术 |
| Language | 中文 |
| Document Type | 学位论文 |
| Identifier | http://ir.xjlas.org/handle/365004/15348 |
| Collection | 中国科学院新疆生态与地理研究所 研究系统 |
| Affiliation | 中国科学院新疆生态与地理研究所 |
| First Author Affilication | 中国科学院新疆生态与地理研究所 |
| Recommended Citation GB/T 7714 | 马秀梅. 基于地面实测光谱的矿区 Cu 元素定量反演研究[D]. 北京. 中国科学院大学,2019. |
| Files in This Item: | There are no files associated with this item. | |||||
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment