蒋夕平,于瀚文,吴芳,黄俊杰.应用PCA和ICA法对岩心光谱的定量解译[J].地质找矿论丛,2016,31(1):121-126
应用PCA和ICA法对岩心光谱的定量解译
The application of PCA and ICA methods in quantitative interpretation of mineral hyper-spectrum
投稿时间:2014-12-12  
DOI:10.6053/j.issn.1001-1412.2016.01.016
中文关键词:  PCA  ICA  岩心光谱  混合像元分离  丰度反演
英文关键词:PCA  ICA  core spectra  separation of mixed pixel  abundance inversion
基金项目:国家重大科学仪器设备开发专项"岩心光谱扫描仪研发与产业化"(编号:2012YQ050250)和南京农业大学基本科研业务费专项资金(编号:KYZ201425)资助。
作者单位E-mail
蒋夕平 南京农业大学理学院, 南京 210095  
于瀚文 南京炮兵学院, 南京 211132  
吴芳 南京农业大学理学院, 南京 210095 wufang318@njau.edu.cn 
黄俊杰 南京地质矿产研究所, 南京 210016  
摘要点击次数: 1400
全文下载次数: 1155
中文摘要:
      岩心光谱属于混合像元光谱,维度多、数据量大,在端元数量、端元光谱及混合矩阵未知的情况下,定量解译岩心光谱以提取岩心所含矿物信息难度大,噪声的存在使问题更加复杂。文章应用PCA和ICA法定量解译岩心光谱主要有三步:采用PCA法预处理混合像元光谱矩阵,在新的特征空间中保留特征值较大的少量特征矢量,有效滤除能量较小的成分,只保留主要成分信息,同时滤除数据中的噪声;采用ICA法分离岩心混合像元,得到混合像元中的端元光谱集,通过矿物识别获取岩心矿物成分;对端元光谱进行归一化处理,基于线性光谱混合模型进行丰度反演,得到岩心混合像元中各端元丰度。通过研究仿真数据获得光谱定量解译的一般规律,创建有效的算法模型,再处理实际测量的岩心光谱,获得了比较理想的结果。
英文摘要:
      The core hyper-spectrum is of mixed pixels with multi-dimensions and massive data. The quantitative interpretation of hyper-spectrum is extremely difficult when the end number of pixel, end spectrum and the mixing matrix are all unknown. If noise appears it is much complicated. with the application of PCA and ICA methods a three-step process is developed for the quantitative interpretation. Firstly PCA method is used to preprocess the pixel spectrum matrix and in new characteristic space several characteristic vectors with larger eigenvalues are kept and simultaneously noise filtered, Secondly, ICA method is used to un-mix the pixels, obtaining collection of end spectrum and identifying mineral components in the core. Finally, the end spectrum is normalized and abundance inversion carried out and each end abundance obtained. The quantitative interpretation law of hyper-spectrum is resulted from study on the simulation data. Algorithm model is built based on the law and the measured spectral of core processed by the model with ideal results.
查看全文  查看/发表评论  下载PDF阅读器
关闭