基于高光谱图像的矿物谱带信息抽取与填图系统

发布时间：2018-05-02 20:52

本文选题：高光谱遥感 + 特征提取　；参考：《南京理工大学》2017年硕士论文

【摘要】：作为一种细分光谱成像技术,高光谱遥感是地质勘探、地物覆盖分析和矿产资源调查的高效的手段。高光谱矿物识别与填图是遥感地质填图由岩性填图到矿物填图的最重要一步,其中矿物光谱特征参量提取与矿物识别的种类和精度是核心。而高光谱图像是图谱合一的数据,需要研究快速并行的高效处理方法。本论文通过设计蚀变矿物分层识别谱带参量化特征提取方法的基础上,构建蚀变矿物识别的规则库,建立推理机,研制一款高光谱图像特征提取和矿物识别系统。主要研究成果如下:(1)提出了一种局部去包络线和多项式拟合的参量化光谱吸收特征提取方法。该方法通过局部去包络线处理,消除背景影响,增强光谱的吸收特征;利用局部多项式拟合,克服光谱噪声影响。在此基础上,该方法针对Al-OH键矿物和Mg-OH键矿物等,设计了组合吸收位置、吸收深度和半谷宽度等参量化光谱特征提取方法。实验表明,该方法光谱特征提取精度高,有效克服了背景和噪声的影响。(2)组合主要的吸收谱带和次要吸收谱带的参量化特征,构建了含Al-OH键矿物和Mg-OH键矿物等的矿物光谱识别规则库;从而设计了基于专家系统的蚀变矿物推理识别方案,并给出了填图方法。(3)研发了一款高光谱图像蚀变矿物特征提取与识别软件。该软件基于多核计算架构,集成局部去包络线和多项式拟合光谱曲线预处理、参量化特征提取、矿物推理识别模块等。经过数据测试,验证了所属Al-OH键矿物和Mg-OH键矿物等矿物识别的类别和精度都大大提升。
[Abstract]:As a subdivision spectral imaging technique, hyperspectral remote sensing is an efficient means for geological exploration, ground cover analysis and mineral resources investigation. Hyperspectral mineral identification and mapping is the most important step from lithologic mapping to mineral mapping in remote sensing geological mapping, in which the types and accuracy of mineral spectral characteristic parameters extraction and mineral identification are the core. Hyperspectral image is a kind of data, which needs to be processed quickly and in parallel. On the basis of designing the method of extracting spectral band parameter quantization feature of altered mineral stratified recognition, the rule base of alteration mineral recognition is constructed, the inference machine is established, and a hyperspectral image feature extraction and mineral recognition system is developed in this paper. The main research results are as follows: (1) A method of extracting the spectral absorption characteristics of parametric quantization based on local de-envelope and polynomial fitting is proposed. In this method, the background effect is eliminated and the absorption characteristic of the spectrum is enhanced by means of local decontamination, and the influence of spectral noise is overcome by using local polynomial fitting. On this basis, the combined absorption position, absorption depth and half-valley width of the combined absorption location, absorption depth and half-valley width are extracted from the Al-OH bond minerals and the Mg-OH bond minerals. Experimental results show that the proposed method has high precision in spectral feature extraction and effectively overcomes the parameterized features of the main absorption bands and secondary absorption bands combined with background and noise. The spectral recognition rule base of minerals containing Al-OH bond and Mg-OH bond is constructed, and the inference and recognition scheme of altered minerals based on expert system is designed. A software for feature extraction and recognition of altered minerals in hyperspectral images is developed. The software is based on multi-core computing architecture, integrates local de-envelope and polynomial fitting spectral curve preprocessing, parameterized feature extraction, mineral inference recognition module and so on. The data tests show that the classification and accuracy of Al-OH bond minerals and Mg-OH bond minerals are greatly improved.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP751

【相似文献】