Research Of Seismic Noise Attenuation Based On Sparse Representation

In the process of oil and natural gas exploration,seismic exploration is a very effective method.The reserves and distribution of underground oil and natural gas can be judged by analysing the obtained seismic records.However,in recent years,with the improved complexity of the environment and geological conditions in the areas explored,the seismic records are becoming more and more difficult to analyze.For example,in the process of microseismic exploration,the received microseismic signals are weak,and the surrounding noise has a great influence on the recognition of the signals.In the desert seismic exploration,there will be a lot of low frequency noise,the valid signal may be flooded by the low frequency signal which has large amplitude.In the face of these problems,we need to take corresponding denoising methods for different seismic signals in order to facilitate the further analysis of seismic data.The main content of this paper is the random noise attenuation in borehole microseismic signal and the low frequency noise and Gauss white noise attenuation in desert seismic signal.Because the seismic signal has strong sparsity and the noise is not sparse,we can attenuate noise in seismic records based on the sparsity of seismic signal.In this paper,we use time alignment shearlet transform to attenuate the noise in borehole microseismic.Shearlet is an effective multi-scale and multi-directional wavelet transform.It can represent the seismic signal sparsely in the transform domain.We can use threshold shrinking in the transform domain and then transform it back to the time domain.In this way,it can remove the noise in two-dimensional seismic data.But when using this method to remove the noise in borehole microseismic signal,the pseudo Gibbs effect will be produced.In order to eliminate the pseudo Gibbs effect,the time alignment shearlet denoising method is proposed in this paper.We use time alignment to remove time delay in seismic traces.The time alignment is done using cross-correlation.After time alignment the event in seismic data will become a horizontal one,the singularities of seismic signal in the direction of space is eliminated,and the seismic signals are more sparse in transform domain.The pseudo Gibbs effect can be reduced when we use shearlet denoising on the seismic data which has horizontal event.The low frequency noise and Gauss white noise attenuation in desert seismic signal are then studied in this paper.Some denoising method can not attenuate the noise in desert seismic signal because that it has both low frequency noise and Gauss white noise in desert seismic signal.Compound sparse denoising(CSD)method isproposed to simultaneously attenuate the low frequency noise and Gauss white noise in desert seismic signal.Compound sparse denoising is an improved method of total variation.Compound sparse denoising adds constraint to the sparsity of seismic signals,and combines with low-pass filter,it can attenuate Gauss white noise and low frequency noise at one time.We use both simulated seismic data and real seismic data to test the methods we proposed in this paper,the experimental results indicate that the proposed methods achieve the desired effect and can improve the signal to noise ratio of seismic data effectively.