The Research Of Transient Electromagnetic Data Denoising Based On Wavelet Transform And Independent Component Analysis

Transient Electromagnetic Method(TEM) is a kind of electromagnetic prospecting Method in time domain.For its unique advantages such as great exploration depth,traversing high resistance layer easily,little affected of the topography,etc.It has been widely applied in the field of source exploration, engineering survey,and environment monitoring,etc. But transient electromagnetic signal belongs to be the wide band signal which the early signal energy decay quickly and the late signal energy decay slowly and is very weak,resulting in the late measured data tend to be various noise interference and influence resolution and accuracy of deep data and causing many difficulties for data processing and inversion interpretation of later period.Although the Signal to Noise Ratio(SNR) can be improved by multiple stacking,increasing the transmission power and improving the sensitivity of the instrument,the actual operation and technology are limited at present.Therefore,seeking for suitable denoising methods to suppress the noise becomes a problem to be solved at once.Traditional denoising methods mainly construct a low-pass filter for denoising by Fourier transform,but Fourier transform in frequency domain analysis can not give transformation situation of some point and spectrum information not given in the time domain analysis, and cause the time-frequency localization contradiction.Wavelet transform which is a method of time-frequency localization analysis can effectively distinguish between mutations parts of the signal and noise signal,so it can realize the signal denoising.Independent component analysis(ICA) can split multi-source mixed signal according to the principle of mutual statistical independence,which can realize the separation of the useful signal and noise signal.This paper is on the basis of predecessors’ work and then systematically study two main denoising theories:the wavelet transform and independent component analysis(ICA) theory.Using these two methods analyzes and compares the validity for the numerical simulation by making programs.Finally it is applied to the measured transient electromagnetic data,and can obtain high quality of the signal-to-noise ratio and achieve the goal of denoising.This paper’s main work and research results are as follows:(1)Introduce the background and the meaning of this paper and analysis the limitation of traditional filter methods.Deeply introduce the research status of signal denoising with wavelet analysis and independent component analysis at home and abroad.Finally expound the main work and the obtained research achivement of this paper.(2)Summarize some characteristics of transient electromagnetic signal and analyze the source of the TEM noise. We realize that the attenuation characteristics of early to late have big difference and it’s frequency range is very wide and many types of electromagnetic noise spectrum and TEM signal spectrum can be overlapped,leading to the limitation of using traditional denoising method.Therefore,putting forward the methods of wavelet analysis and the ICA into TEM signal extraction.(3)The signal denoising based on wavelet transform. First study the basic theory of wavelet analysis systematically,including the introduction of wavelet transform and its classification,the multiresolution analysis,Mallat algorithm and commonly used wavelet functions,etc.Then introduce the wavelet threshold method,the wavelet modulus maxima method and wavelet packet denoising algorithm,and make algorithm programs to verify denoising effects of the three methods by simulation experiment and analysis effectiveness of the simulation results. Experimental results show that the effectiveness of the three methods have been well verified.(4)The signal denoising based on independent component analysis. First,the basic theory of ICA is introduced briefly,including the ICA model,basic assumptions and uncertainty problems,etc.Then we focuses on two kinds of blind source separation technologies which are FastICA and natural gradient learning method,and makes algorithm programs to verify the effectiveness of the two methods by simulation experiment and analysises the result of the experiment at the same time. Experiments show that the two methods have good effect for signal separation.(5)The denoising of transient electromagnetic data.First of all,we introduces the signal noise model,the denoising principle and the signal quality assessment standards. And then we apply wavelet analysis and independent component analysis to denoising analysis for TEM simulated data,and analysis the advantages and disadvantages of these methods respectively. Then comparing and analysising the denoising effect of the wavelet analysis and FastICA method for different noise intensity of signal denoising. We find FastICA which is more robust than wavelet method and is more suitable for dealing with strong noise signal and then put forward a kind of deboising cmbined algorithm based on FastICA and wavelet analysis. Finally we apply this joint algorithm to the measured transient electromagnetic data denoising, and good results have been achieved and it can be extended to TEM data preprocessing.(6)Summarize the research achievements and some advice is given to the existing problems of this paper,and discuss the future development of transient electromagnetic signal denoising theory.