Research On Key Technology Of High Resolution Time-frequency Analysis In Seismic Processing

With the rapid progress of oil and gas exploration,deep exploration,complex geological bodies,thin layers,thin interbeds,and small-scale reservoirs have become the hot topics in area of exploration and development.The frequency will be absorbed and attenuated as the seismic wave propagates downward.As thus,time-frequency analysis technology can describe the relationship between seismic signal's frequency and time,and then carry out fluid identification,prediction of reservoir thickness and so on,which have great significance.Therefore,how to improve time-frequency resolution of seismic data,identify effective information at low SNR in time-frequency domain,and identify weak signals in time-frequency domain are the main issues to be studied in this paper.In view of that above,this paper improves or proposes some new time-frequency analysis methods,setting out from linear time-frequency analysis,quadratic time-frequency analysis,and curvelet transform,respectively.The linear time-frequency analysis method is the most commonly used type among traditional time-frequency analysis methods.Due to the influence of the uncertainty principle,it is not possible to achieve high time-frequency resolution in both time domain and frequency domain.Synchrosqueezing transform can compress time-frequency energy and reassign it to the time-frequency ridge,and improve the resolution in the time-frequency domain.Foreign scholars have proposed a second-order synchrosqueezing transform method with higher resolution.But the forms are not concordant,which is not conducive to the popularization and application.This paper derives second-order synchrosqueezing short-time Fourier transform and second-order simultaneous synchrosqueezing wavelet transform in time domain and frequency domain respectively and then unifies the two forms.Besides,we apply the second-order synchrosqueezing wavelet transform to actual seismic data processing to identify favorable traps.The time-frequency spreading situation of effective signal and random noise on second-order synchrosqueezing Gabor transform and second-order synchrosqueezing wavelet transform are different.And then we can suppress the random noise and preserve the effective information by local similarity algorithm.The curvelet transform,which is an extension of the wavelet concept,is often used to perform denoising,data reconstruction,and attribute extraction etc.In this paper,a time-frequency analysis method based on the curvelet transform is proposed.We build a mother curvelet that adapts to seismic data and the rotation mode is shearing rotation,use the optimal angle curvelet coefficient to extract time-frequency information,and apply the idea of synchrosqueezing wavelet transform into time-frequency spectrum of curvelet transform to improve the resolution in time-frequency domain.The traditional time-frequency analysis method is aimed at one-dimensional signal processing,and it does not take the lateral variation of signal into account.For data with a low SNR,it cannot identify the valid information accurately.Although two-dimensional time-frequency methods have been proposed before,the time-frequency data obtained by this way has large redundancy,low computational efficiency,and high memory usage.Therefore,it has not been widely used.The time-frequency analysis method based on curvelet transform can not only improve these matters,but also identify effective signals at low SNR more accurately.The representative of the quadratic time-frequency analysis method is the Wigner distribution which has a very high time-frequency resolution.However,due to the interference of cross terms,it is less widely used compared to the linear time-frequency analysis methods.There are many methods which can suppress the cross terms,but they all deal with the time-frequency energy of the Wigner distribution,such as the cross-term suppression methods represented by the smoothed pseudo-Wigner distribution.This kind of time-frequency method is not effective for weak signal recognition.This paper proposes a Wigner pseudo-phase spectrum time-frequency distribution method,which can construct Wigner phase spectrum by the Hilbert transform of Wigner distribution,and then obtain its time-frequency ridge or carry out filter processing to reduce the interfering cross-terms,and finally get a new time-frequency spectrum.Traditional time-frequency analysis methods have the problems of low resolution,difficult identify of useful information in low SNR,and difficult extractation of weak signals.Based on these problems,this paper derives second-order synchrosqueezing short-time Fourier transform and second-order synchrosqueezing wavelet transform in time domain and frequency domain respectively,and then improve the time-frequency resolution of seismic data;proposes a time-frequency analysis method based on curvelet transform,which has strong signal recognition ability for low SNR data;proposes a Wigner pseudo-phase spectrum time-frequency distribution method,it can identify weak signals effectively under the premise of ensuring higher time-frequency resolution.Both model tests and actual data processing verify the validity and reliability of the proposed method in this paper.