【作者】 何中海；

【导师】 何彬彬；

【作者基本信息】 电子科技大学 ， 地图制图学与地理信息工程， 2011， 硕士

【摘要】 自20世纪80年代以来,高光谱遥感已成为遥感发展的重要方向。高光谱遥感在地质中的应用更是其最重要也是最为成功的领域之一。地理环境恶劣,交通不便的我国西部高寒山区,蕴含了大量矿产资源,利用高光谱遥感进行地质资源调查是一种便利和高效的勘探手段。但是,由于岩矿光谱易受环境影响,在不同地区会产生一定的变异,这给岩矿填图带来了不确定性。建立高寒山区岩矿波谱库、分析岩矿光谱变异性以及寻找到合适的岩矿填图方法便显得十分重要。论文以野外波谱测试、分析和岩矿波谱特征变化研究为基础,建立较为适合的岩矿信息提取方法,并利用Hyperion高光谱遥感数据在青海东昆仑重点地区进行了岩矿填图实验。主要研究工作和成果包括:(1)对高寒山区的岩矿波谱机理研究。通过野外光谱采集和实地勘察,研究和总结了研究区典型岩矿的波谱特征形成机理及其在高寒山区的自然条件下产生的变异。研究的岩石包括砂岩、花岗岩和闪长岩,矿物主要是蚀变铁矿和粘土矿物。通过实地考察和光谱分析,对岩矿的波谱特征稳定性做了总结,并为下一步的矿物填图做好指导工作。(2)针对相似性矿物难以区分的问题,论文提出了三种新的岩矿光谱匹配方法:权重光谱角填图、耦合整体光谱匹配和局部光谱匹配、特征参数匹配,并应用最为广泛的美国内华达州铜矿试验区对三种方法进行了技术验证。同时,采用Hyperion数据在东昆仑地区进行了矿物识别实验,取得了较好的效果。实验证明三种新的方法在对相似性矿物的区分能力上,较传统的光谱角制图、光谱特征拟合等方法有所提高。(3)在东昆仑五龙沟、石灰沟进行了岩性填图。岩性填图采用了三种流程。一是基于图像端元提取的岩性识别,它是根据MNF变换,像元纯度指数和n维可视化的手段直接从图像中提取纯端元作为参考波谱进行岩性识别;二是根据野外测试光谱进行岩矿填图,它是直接从野外测试光谱中选择需要识别的参考波谱与校正后的反射率数据进行匹配填图;最后是利用实验室波谱库中的岩矿波谱进行填图。实验结果证明,三种岩性填图流程中,基于图像端元提取的岩性填图易于得到图像的丰度信息,基于野外实测光谱的岩性填图则相对简单,而基于实验室波谱的岩性填图效果较差,主要是因为高寒山区的部分岩石发生了粘土化和其他变化;光谱匹配方法中,在地物信息强烈的情况下,常用的光谱角制图和光谱特征拟合的效果都不错,特别是光谱角制图更具优势,而当地物信息较弱,或者在背景和噪声影响下难以识别时,利用自适应一致估计法,将目标从背景和噪声中分离出来,取得的效果最佳。更多还原

【Abstract】 Since the 1980s, hyperspectral remote sensing has become the important direction of remote sensing. Its application in geological mapping is the one of the most important and successful field. In western high and cold mountains, there are a number of minerals, but the geographical environment is bad and the transportation is inconvenient, so the use of hyperspectral remote sensing in geological investigation is a kind of convenient and efficient exploration method. However, under the influence of environment factors, the rocks and minerals spectrums will produce certain variations in different areas of the world, which could bring adverse effect to the rocks and minerals mapping. Establish spectral library of rocks and minerals in high and cold mountains, analysis spectral variability and find appropriate mapping methods are very important.This paper has set up the proper information extraction methods and carried on rocks and minerals mapping with the Hyperion hyperspectral data in the east Kunlun based on the field spectral testing, analysis and its variability. The main works and results include:(1) This paper has made the research on the rocks and minerals spectral mechanism in high and cold mountainous. Through the field investigation and field spectral colletion, this article has researched and summarized the spectral characteristics, spectral mechanism and the spectral variations of typical rocks and minerals in the high and cold region. The rocks studied in this article include the sandstone, granite and diorite, and the minerals include the alteration iron ores and the clay minerals. Through field investigation and the spectral analysis, this article has made the summary to the stability of rocks and minerals spectral characteristics, and taken it as the preparative work for the mapping. (2) In order to distinguish similarity minerals, this paper proposed three new recognition methods: the weight spectral angle mapping method, the mapping method combining overall with partial spectral matching and the feature parameters matching method. This paper utilized the most widely used American Nevada cuprite experimental plot to confirm these methods, then carried on the minerals mapping in east Kunlun area with the Hyperion data and obtained the good results. The?experimental results certify that three kinds of new methods have the stronger discerniblity ability to similarity minerals than traditional SAM (Spectral Angle mapper) and SFF (Spectral Feature Fitting).(3) This artcle has carried on the rocks and minerals mapping in Shihuigou and Wulonggou of East Kunlun. The lothological mapping took three kinds of flows. One is the lothological recognition flow based on the image endmember extraction, which takes the MNF transform, the pixel purity index and the n-Dimensional Visualization to extract the endmembers as the reference spectrums from the image; the second kind of flow is to carry on the rocks and minerals mapping according to the spectrums collected in the field which directly distinguishes the object by matching the field spectrum and the image reflectance spectrum; the last kind of flow is to carry on the rocks and minerals mapping according to the library spectrums provided by USGS and JHU, et al. The experiment results show that the lithologic mapping based on image endmember extraction is easy to get rocks abundance information and the lithologic mapping based on the field measured spectrums is relatively simple, but the lithological mapping based on laboratory spectrums is poor, mainly because of the rocks spectrums have changed in high and cold mountains; In the process of spectrum matching, if the target information is strong, the effect of SAM and SFF would be good, especially the SAM was more advantage; while target information is weak, or influenced by the background and noise, the ACE(Adaptive Coherence Estimator) could separate the target from background and noise well.更多还原