Research And Application Of Improving Resolution Of Seismic Signal Based On Compressed Sensing

In recent years,high signal-to-noise ratio and high-resolution seismic data have been the goal of the oil and gas exploration industry.However,the acquired seismic data are affected by various aspects,resulting in low signal-to-noise ratio and insufficient resolution,which seriously affects the interpretation of subsequent seismic data.Therefore,the reasonable and effective processing can provide effective data guarantee for reservoir prediction,oil and gas detection and other work.This paper will focus on enhancing the resolution and signal-to-noise ratio of seismic data,and apply the theory of compression perception to seismic data processing.The research content includes the following points:(1)Most of the denoising methods based on the compressed sensing theory are based on the sparse decomposition principle,which can remove the noise by detecting the noise signal in the complete atomic database.In other words,it is to reconstruct the original signal with a larger coefficient and cover the noise with a smaller coefficient to eliminate the noise.In this paper,two kinds of algorithms,namely,the Orthogonal Matching Pursuit algorithm(OMP)and the Fast Iterative Soft-Thresholding Algorithm(FISTA),are introduced respectively,and their denoising effects under the wavelet matrix and the K-SVD dictionary matrix.They are based on the sparse algorithm to denoise the seismic data and reconstruct the original seismic signal.In sparse representation part,OMP and FISTA algorithm mostly use sparse transform such as Wavelet,Fourier and Curvelet.These sparse bases are fixed and the fixed sparse basis can't represent all the signals accurately and can't remove the noise better.The KSVD dictionary learning can match the signal structure better according to the adaptive training dictionary of the sparse characteristics of the signal itself.In addition,after updating the dictionary,K-SVD dictionary learning can make the atom operate on the atom in a more simple and effective way,avoiding the matrix inverse operation,and updating the current atom and the related coefficient to accelerate the learning process of the dictionary.The results of theoretical model and practical data processing show that the K-SVD dictionary learning algorithm can effectively improve the signal-tonoise ratio of theoretical model and actual seismic data,the reconstruction effect is more prominent than the algorithms under the wavelet sparse basis,and can accurately construct the original theoretical model and actual seismic data.(2)There are usually missing channels in the actual seismic data,so the signal-tonoise ratio of incomplete data is relatively low.In order to fully recover the missing seismic data information and improve the signal-to-noise ratio of seismic data.In this paper,based on the finite seismic data and the interpolation algorithm of compressed sensing theory,the interpolation reconstruction of missing seismic traces by POCS algorithm is mainly introduced,and two algorithms,IST and FISTA,are compared.The theoretical model and the actual data processing results show that POCS algorithm has good effect,high precision reconstruction and other characteristics,and has high calculation efficiency,which can reasonably and effectively restore the original channel information missing from seismic data.(3)The low-frequency compensation method based on compression sensing theory is often used in seismic data with low-frequency signal missing and difficult to accurately describe reservoir information.Compared with the high-frequency component,the low-frequency information has stronger penetration,longer propagation and slower attenuation,so the information contained in the low-frequency is more reliable than the high-frequency information,and can reflect the formation information better.The lack of low-frequency will lead to the unreliable seismic data information collected,which will cause the false high-resolution phenomenon on the seismic profile.In this paper,a new compression sensing algorithm is proposed to fit the seismic data in full frequency band,and the low frequency part of the fitting results is compensated to the seismic data.After low-frequency compensation,the seismic section's stratification and in-phase axis continuity are improved,which is more consistent with well curve,and the resolution of seismic data is effectively improved,which is of great significance for reservoir prediction and recognition.In this thesis,firstly,the denoising method based on the theory of compressed sensing is used to denoise the seismic data,then the missing seismic channel data is interpolated and reconstructed,and finally the low-frequency compensation processing of the seismic data is carried out by the frequency expansion method based on the theory of compressed sensing,so as to improve the resolution of the seismic signal.To sum up,the theory of compressed sensing has been successfully applied to the processing of seismic data with low SNR and low resolution.The practical results show that the theory of compressed sensing has a good application prospect in improving the signal to noise ratio and resolution of seismic data.