光谱特征提取对近似矿物光谱差异性影响分析

发布时间：2018-03-06 23:16

本文选题：近似矿物　切入点：光谱特征提取　出处：《光谱学与光谱分析》2017年03期 　论文类型：期刊论文

【摘要】：矿物光谱特征是基于光学遥感数据对矿物进行种类识别及定量反演的理论基础,光谱特征提取是高光谱数据常用的技术手段,但在多光谱数据中较少涉及。近似矿物识别是矿物光谱分类应用中的难点,目前还缺少有效指标来指示近似矿物类别光谱的差异性。光谱特征提取有望提高矿物分类精度,但该处理对近似矿物光谱差异性的影响还缺少相关研究。本文从矿物光谱差异性的原理出发,通过类间和类内光谱角的比值体现不同类别群体差异,并引入样本量因素,提出了类别可分比作为近似矿物光谱差异性的指标。以明矾石和高岭石两种近似矿物为例,对USGS光谱库光谱及Hyperion,ASTER,OLI等传感器的模拟数据进行光谱特征提取处理,通过对比处理前后矿物光谱差异性的变化,分析光谱特征提取对近似矿物光谱差异性的影响。实验结果表明,有效的光谱特征提取可以显著提高近似矿物光谱差异性,并且光谱分辨率越高,近似矿物光谱差异性越大。此外,光谱分辨率及中心波长设置对于包络线去除结果有很大影响,多光谱数据吸收特征提取效果有待进一步提高。该研究为今后近似矿物光谱识别精度的提高奠定了基础,也为未来新型遥感找矿传感器参数设置提供了参考。
[Abstract]:The spectral feature of minerals is the theoretical basis for classification identification and quantitative inversion of minerals based on optical remote sensing data. Spectral feature extraction is a common technique for hyperspectral data. But it is less involved in multispectral data. Approximate mineral recognition is a difficult point in mineral spectral classification. At present, there is still a lack of effective indicators to indicate the spectral differences of approximate mineral types. The extraction of spectral features is expected to improve the classification accuracy of minerals. However, the effect of this treatment on the approximate mineral spectral difference is still lack of relevant research. Based on the principle of mineral spectral difference, the ratio of spectral angle between classes and within classes is used to reflect the difference of different groups, and the sample size factor is introduced. Taking alunite and kaolinite as examples, the spectral characteristics of USGS spectra and simulated data of Hyperion Aster Oli sensors are extracted and processed. By comparing the changes of mineral spectral differences before and after treatment, the effects of spectral feature extraction on the approximate mineral spectral differences are analyzed. The experimental results show that the effective spectral feature extraction can significantly improve the approximate mineral spectral differences. And the higher the spectral resolution, the greater the spectral difference of approximate minerals. In addition, the spectral resolution and the setting of central wavelength have great influence on the envelope removal results. The extraction effect of absorption characteristics of multispectral data needs to be further improved. This study lays a foundation for improving the accuracy of near mineral spectral recognition in the future and provides a reference for the parameter setting of the new remote sensing sensor for mineral prospecting in the future.
【作者单位】：中国矿业大学(北京)地球科学与测绘工程学院;中国科学院遥感与数字地球研究所高光谱遥感应用技术研究室;
【基金】：国家自然科学基金项目(41171306) 国家高技术研究发展计划项目(2013AA12A302)资助
【分类号】：P627

【相似文献】