主成分分析在航空瞬变电磁去噪中的应用  

A denoising method based on principal component analysis for airborne transient electromagnetic data

作　　者：武莹[1] 陆从德[2] 杜兴忠[3] 余小东[2] 

机构地区：[1]成都理工大学信息科学与技术学院,成都610059 [2]成都理工大学地球物理学院,成都610059 [3]中国水电顾问集团贵阳勘测设计研究院,贵阳550081

出　　处：《物探化探计算技术》2014年第2期170-176,共7页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家863计划(2013AA063904);四川省教育厅重点项目(11ZA045);成都理工大学中青年骨干教师培养计划(GK201105)

摘　　要：由于航空瞬变电磁二次场数据是一种宽频带信号且幅度较小,因此常常受到随机噪声、天电噪声以及人文噪声的影响。研究针对上述噪声的去噪方法,可以提高处理和解释的精度。这里给出了基于主成分分析的航空瞬变电磁去噪方法,首先应用主成分分析方法对叠加抽道后的单测点衰减曲线进行分解,然后应用曲线趋势对比法和L曲线法对分解后的成分进行分析,确定出有效信号成分,最后使用这些有效信号成分进行重构,从而达到抑制噪声的目的。模拟信号分析以及实测数据的实验结果表明,应用主成分分析法处理航空瞬变电磁数据,在保持反映地下介质分布的有效信号幅度基本不变的条件下,能有效抑制天电噪声和人文噪声,为保幅处理和解释奠定了良好的基础。As a kind of broadband signal, the airborne TEM secondary field data with smaller amplitude are often contaminated by random noise, sferics noise and man made noise, etc. How to seek the denoising methods with good performance for the a bove several noises has become one of key parts to improve the accuracy of processing and interpretation. This paper presents a denoising method based on principal component analysis （PCA）. Firstly, PCA is applied to decompose the single survey point data which has been stacked and sampled. Then, the components of the effective signal are discriminated by L curve method and the comparison method of curve trend. At last, these effective signal components are reconstructed to reduce noises. The experimental results arid analysis on both simulated data and the measured data show that the sferics and the man made noise can be effectively suppressed with maintaining the amplitude of effective signal which reflect the distribution of subsurface media when the denoising method based on PCA is applied to process airbone TEM. The denoising method proposed shows a route of the amplitude preserved processing and interpretation.

关 键 词：航空瞬变电磁法 去噪 主成分分析 L曲线 信号重构 

分 类 号：P631.326[天文地球—地质矿产勘探]