局域均值分解和小波阈值在大地电磁噪声压制中的应用

Application of local mean decomposition and wavelet threshold in magnetotelluric noise suppression

大地电磁测深法是基于电磁感应原理,利用天然交变电磁场来研究地下岩层的电学性质及其分布特征。然而,天然电磁场频带范围宽、信号微弱,在实际测量中大地电磁信号极易受到各种电磁噪声干扰,严重影响了后续的电磁法反演解释水平。针对这一难题,将局域均值分解(LMD)的自适应性和小波分析的多分辨性相结合,提出基于局域均值分解和小波阈值的大地电磁噪声压制方法。将含噪信号进行LMD分解得到若干阶乘积函数(PF)分量;根据大地电磁信噪特征保留PF_1分量,仅对其余各阶PF分量选取合适的小波阈值进行降噪处理;叠加重构获得大地电磁有用信号。通过计算机模拟典型强干扰,研究不同小波函数、分解层数及阈值方式下算法的去噪性能,并将其应用于矿集区实测大地电磁数据处理。实验结果表明,所提方法能较好地提取出叠加在微弱大地电磁信号上的大尺度强干扰的轮廓特征,视电阻率曲线更为光滑、连续,低频段的大地电磁数据质量得到了明显改善。

The magnetotelluric sounding method is a method based on the principle of electromagnetic induction, to detect electrical properties and distribution characteristics of underground rock layers by use of the natural alternating electromagnetic field. However, the natural electromagnetic field has a wider frequency band range and weak signals, and these signals are easy to be disturbed by all kinds of electromagnetic noise in the actual measurement, so the interpretation level of the subsequent electromagnetic inversion method is seriously effected. In order to solve this problem, combining the adaptability of local mean decomposition （LMD） with the multi-resolution of wavelet, here the magnetotelluric noise suppression method based on local mean decomposition and wavelet threshold was proposed. First of all, LMD was used to divide a noisy signal into a number of product functiion （PF） components. Then, according to magnetotelluric signal- to- noise characters, PFt component was retained, the appropriate wavelet threshold was chosen to denoise the resteach PF component. Finally, the magnetotelluric useful signal was obtained with superposition and reconstruction. Using a computer to simulate typical strong interferences, the denoising performances of the proposed method were studied under the conditions of different wavelet functions, decomposition layers and threshold modes, and the method was applied to process the magnetotelluric data measured in ore concentration areas. The results showed that the proposed method can better extract the outline features of large-scale strong interferences superimposed on weak magnetotelluric signals, and the apparent resistivity curve is more smooth and continuous ; the quality of magnetotelluric data is improved significantly in lower frequency bands.