Desert Seismic Signals Processing Based On Low-Rank Extraction In Time-frequency Domain

Nowadays,the demand for nonrenewable resources such as oil and natural gas is increasing with the development of human society,while the oil-gas resources in conventional hydrocarbon reservoir areas and easy to exploit areas are constantly consumed.Therefore,the research on oil-gas exploration in hydrocarbon reservoir areas with complex environment and difficult to exploit has become a hot issue in the field of resource exploration.The desert area in Tarim Basin is rich in oil-gas resources,so it is of great significance to carry out geological exploration,analyze geological structure and explore resource reserve.As a high-precision and low-cost resource exploration method,seismic exploration method can provide a set of detailed and more accurate stratigraphic structure data from shallow to deep.However,due to the particularity and complexity of the environment in the desert area,the collected seismic data are often affected by the random noise of the desert,which is non-Gaussian,nonstationary and low frequency,the effective signal is submerged seriously.It is difficult to achieve the "three high" requirements of high signal-to-noise ratio,high resolution and high fidelity of seismic data,and the seismic exploration data with high quality cannot be obtained.Therefore,it is of great necessary for oil-gas exploration to study the method of suppressing desert seismic random noise effectively.Up to now,many excellent scholars have proposed,improved and applied many effective algorithms.In the traditional method of suppressing seismic noise,the bandpass filter,wavelet transform and threshold have been widely used,and have achieved good processing effect.However,due to the frequency overlap of effective signal and low-frequency random noise in desert seismic data,the upper rule will be limited to some extent.In recent years,time-frequency decomposition method and low-rank matrix approximation algorithm have developed rapidly.Time-frequency transform(TFT)is used to localize individual oscillatory components of a recorded signals.Because of the non-stationary characteristics of seismic signals,it is often more informative to use their time-frequency representation instead of considering the signal in the frequency or time domain(one-dimensional).In the time-frequency domain,the low-frequency desert seismic noise can be regarded as retaining the low-rank characteristic,and its rank in the transform domain is lower than that of the effective signal in the transform domain.Therefore,the low-rank component extraction method can be utilized to suppress the noise.This paper is based on time-frequency decomposition method and low-rank matrix approximation algorithm for desert seismic signal denoising algorithm.According to the difference of matrix rank between effective signal and noise of desert seismic data in time-frequency domain,this paper proposes a method of extracting low-rank components in time-frequency domain to realize signal-noise separation.Firstly,VMD is used to process the seismic signal in a single trace,which is decomposed into several modes.Without discarding any modes,the decomposed modes are recombined to construct a signal component matrix,and then the low-rank components are extracted by using OptShrink.Finally,the purpose of denoising is achieved.This method is called VMD-OptShrink.However,due to the characteristic of OptShrink itself,it is difficult to select rank value.Therefore,based on this method,this paper proposes a method to simplify rank selection in time-frequency domain,which is called VMD-SSGoDec.The difference between the two methods is that the number of low-rank extraction times in the whole algorithm is different.VMD-SSGoDec simplifies rank selection by reducing the dimension of constructed signal component matrix Rank selection to achieve noise suppression.In this paper,the above methods are applied to synthetic desert seismic records and real desert seismic records to verify the effectiveness of the algorithm proposed in this paper.The denoising performance of this method is proved from time-domain,frequency-domain,SNR enhancement,mean square error calculation and other aspects,and compared with the classical methods proposed.The experimental results of synthetic record and real record show that the denoising effect of this method is better than that of the previous method in low-frequency noise in desert.The effective signal is kept intact.The resolution and continuity of seismic events are improved obviously,and the suppression of surface wave is also very thorough.In synthetic record,SNR can be increased about 13 dB,which fully shows that the algorithm proposed in this paper is in advantages of desert seismic signal noise suppression.