匹配追踪抑制大地电磁场噪声影响的研究

发布时间：2018-05-30 08:55

本文选题：大地电磁测深 + 匹配追踪法　；参考：《中国地质大学(北京)》2015年硕士论文

【摘要】：大地电磁测深法是近年来最成熟的电法勘探技术之一,是研究地质电性构造、矿产电法勘探以及工程电法勘探的重要手段。由于其频带宽信号弱,易受日益广泛的自然人文环境噪声的干扰,因此有效的压制噪声,提高大地电磁数据的信噪比,保证大地电磁数据质量,始终是大地电磁数据采集与数据处理的核心问题。然而大地电磁信号具有非线性、非平稳和非最小相位的特点,不满足Fourier变换的条件,因此为了大地电磁测深信号噪声压制引入了短时Fourier变换、小波变换、Hilbert-Huang变换以及S变换等方法进行去噪,对时间域定位和频率域做诸多改进。本文引进一种新的信号分解方法,匹配追踪法,通过建立冗余的过完备波形原子库,实现贪婪算法分解步骤,对大地电磁信号分解得到线性表出,并利用Wigner-Ville谱得到大地电磁信号的时频分布图,认识大地电磁测深信号受强噪声干扰的时频谱特征,通过加载理论噪声,理论信号以及噪声在时频图上的特征,认识噪声分离的规律,并通过对表出原子的特征进行分析来达到信噪分离,而最终达到大地电磁测深受强噪声干扰信号去噪的目的。大地电磁测深信号噪声干扰按照形态进行分类,分为脉冲噪声、谐波噪声、三角噪声、似充放电噪声、阶跃噪声以及方波噪声,为去噪提供含噪声信号模型;介绍匹配追踪基本理论,利用高斯窗函数尺度变换、时移、频移和相位变化构造Gabor过完备原子库;采用贪婪算法把理论模型函数以及大地电磁测深模型信号分别加载不同种类噪声,在过完备库上分解并且重构,验证匹配追踪法分解重构能够在忽略残差的情况下恢复信号,为信噪分离打下基础;引入Wigner-Ville分布对重构信号进行时频谱分析认识各种噪声在时频谱上的特征,为信噪分离提供依据。通过时频谱分析理论模型函数加载噪声以及大地电磁测深模型加载信号的处理取得的信噪分离规律,应用于实际大地电磁测深数据受各类噪声影响信号的去噪处理,为大地电磁测深信号的噪声抑制提供新方法和思路。
[Abstract]:Magnetotelluric sounding (MT) is one of the most mature electrical exploration techniques in recent years. It is an important means to study geo-electric structure, mineral electrical exploration and engineering electrical exploration. Because the frequency band signal is weak and easily disturbed by the increasingly widespread natural and human environment noise, the noise is suppressed effectively, the signal-to-noise ratio of magnetotelluric data is improved, and the quality of magnetotelluric data is guaranteed. It is always the core problem of magnetotelluric data acquisition and data processing. However, magnetotelluric signals have the characteristics of nonlinear, non-stationary and non-minimum phase, which do not meet the Fourier transform condition. Therefore, a short-time Fourier transform is introduced to suppress the noise of magnetotelluric sounding signals. Wavelet transform Hilbert-Huang transform and S transform are used to de-noising and many improvements are made in time domain localization and frequency domain. In this paper, a new signal decomposition method, matching tracing method, is introduced. By establishing a redundant over complete waveform atomic library, the greedy algorithm is implemented to decompose the magnetotelluric signal, and a linear representation is obtained for the decomposition of the magnetotelluric signal. The time-frequency distribution diagram of magnetotelluric signal is obtained by using Wigner-Ville spectrum, and the time-frequency characteristic of magnetotelluric sounding signal disturbed by strong noise is recognized, and the characteristics of theoretical noise, theoretical signal and noise on time-frequency chart are loaded by loading theoretical noise, theoretical signal and noise. The law of noise separation is recognized, and the signal noise separation is achieved by analyzing the characteristics of the atoms. Finally, the purpose of de-noising the intensely noisy signals in magnetotelluric measurement is achieved. The noise interference of magnetotelluric sounding signal is classified into pulse noise, harmonic noise, triangular noise, quasi-charge-discharge noise, step noise and square wave noise. The basic theory of matching tracing is introduced. The Gabor over complete atomic library is constructed by using Gao Si window function scale transformation, time shift, frequency shift and phase change. The greedy algorithm is used to load the theoretical model function and magnetotelluric sounding model signal into different kinds of noise, decompose and reconstruct the signal on the over-complete database, and verify that the matching tracing method can recover the signal under the condition of neglecting the residual error. It provides the basis for the separation of signal and noise, and introduces the Wigner-Ville distribution to analyze the time spectrum of reconstructed signal to understand the characteristics of various noises in the time spectrum, which provides the basis for the separation of signal and noise. The signal and noise separation rules obtained from the signal processing of magnetotelluric sounding model and magnetotelluric sounding model are applied to the de-noising processing of the actual magnetotelluric sounding data affected by various kinds of noise, which is obtained by time-spectrum analysis theory model function loading noise and magnetotelluric sounding model loading signal processing. It provides a new method and train of thought for noise suppression of magnetotelluric sounding signal.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.325

【参考文献】