Study On The 3D Deghosting And Migration Of Ghost By Sparse Constraint

In marine seismic data acquisition,ghost is an inevitable problem.Since the airwater interface is a strong reflector,the acoutstic wave is reflected downward when it propagates to the interface,and received wavefield consists of the upgoing primaries and downgoing ghost.The ghost distorts the waveform of upgoing primary waveform,causes a notch effect,which damaging the frequency information and reducing the longitudinal resolution of the seismic data.In traditional conception of marine seismic data processing,ghost is considered as a kind of coherent noise,which needs to be suppressed before migration imaging.With the technological innovation and cognitive improvement,people realize that the “noise”(e.g.multiple)in seismic processing contains rich information of underground media.If it is used in the migration process,it can improve the imaging quality and reduce the difficulty of subsequent inversion and interpretation.In recent years,people have succeed in utilizing multiple as a supplement to the conventional primary based migration to expand the illumination range to solve complex structural imaging problems,such as steep dip angle and rock coverage.Moreover,the research on ghost utilizing in the image quality improvement has not been carried out.According to observation system of streamer,ghost has a little difference in delay time and waveform,compared with upgoing primaries.And ghost can be regarded as effective information after separation.Therefore,the paper has carried out research on the ghost utilizing in ocean streamers.We start from 2D case and the method of ghost separation for 2D streamer data are studied.Considering the practical application and separation result,the method of multi-shot fast ghost separation and the 3D sparse constrained inversion method in Radon domain are studied.Then the 2D separation algorithm is extended to 3D,and the methods of 3D data regularization and 3D ghost separation are studied.Finally,the potential of ghost utilizing to improve the quality of reverse time migration imaging is studied.Firstly,this paper studies the generation mechanism of ghost,and analyzes the effects of receiver depth,attenuation factor and offset for upgoing primaries.Then we review the deghosting methods in different domains(e.g.frequency domain,frequencywavenumber domain,linear Radon domain),and analyze the advantages and disadvantages of each method.Considering the time efficiency of actual condition,the pseudo-3D ghost delay formula is deduced,and the multi-shot ghost separation strategy is proposed.The Radon domain 3D sparse constrained inversion method is introduced to improve the inversion precision and avoid the local minimum during inversion.Both the synthetic and field data examples verify the effectiveness of the algorithm.The strategy rearranges the 2D multi-shot gathers to a 3D data cube,and utilizes the consistency of the streamer offset information to simplify the calculation of ghost delay time and the multi-shot inversion process.In this way,the simultaneous inversion for multi-shot gathers is achieved.Furthermore,the sparse constrained inversion method is introduced,which mapping the data to the 3D sparse transformed domain.Based on the characteristics of the sparse transform in highlighting coherent events,the inversion is solved in the space with better sparsity.Compared with the traditional 2D ghost separation method,the pseudo 3D ghost separation method can achieve simultaneous inversion for multi-shot gathers,and the calculation time is close to the one shot gather ghost separation,which greatly improves the calculation efficiency.In addition,the information in offset direction is included during inversion,which avoids the inversion process falling into local minimum values and improves the accuracy of the inversion.Secondly,this paper extends the 2D ghost separation method to 3D.We introduce the sparse transform based data regularization method for 3D seismic data and introduce 3D Radon transform in data reconstruction to accelerate the convergence speed and improve the reconstruction result.After obtaining the dense sampling data in crossline direction,the 3D ghost separation method is studied.By proposing the 3D ghost propagation model,the deghosting operator is rewritten as a dominant diagonal matrix cell array,so that the 3D ghost propagation can still be achieved by matrix multiplication.Therefore,we can estimate the ghost-free upgoing primaries by a largescale inversion algorithm.Both the synthetic and field data examples verify the effectiveness of the algorithm.Finally,considering the small difference of delay time in offset direction,a fast 3D ghost separation method based on crest factor maximization is proposed.According to the kinematics of 3D seismic data,a prenormal move out correction(NMO)is adopted,and the 3D parabolic Radon transform is applied to sparsely represent the 3D data.The ghost delay time corresponding to the maximum of crest factor is extracted in the certain search range to construct the ghost separation operator.An updated ghost operator for ghost wave separation is obtained.The method overcomes the disadvantages of the traditional 3D ghost separation method,and is beneficial to the practical application of the 3D ghost separation method.Finally,we study the application prospects of ghost in reverse time migration imaging.By comparing the cross-correlation imaging condition and the excitation amplitude imaging condition,it is recognized that the cross-correlation imaging conditions are widely used and can provide stable imaging results,but the computational efficiency is low,which is not conducive to industrial application;The excitation amplitude imaging conditions are computationally efficient,and the imaging value is related to the reflection coefficient,which can provide a reference for the subsequent migration velocity analysis,etc.However,there are some imaging blurs in some structure.Comprehensive consideration of computational efficiency and imaging results,we propose ghost constrained the excitation amplitude imaging conditions.According to the delay time between ghost and primaries,the stacking range is determined.All the imaging values in the stacking range are superimposed to compensate the problem of the fuzzy imaging.Through the thin-layer model and part of the Marmousi model imaging examples,the imaging quality by the excitation amplitude imaging condition is improved.The modified excitation amplitude imaging conditions balance the imaging quality and computational efficiency,and have more advantages in industrial application.Most of the methods proposed in this paper combine the idea of compressed sensing,which mapping the data to the model space with higher sparsity by sparse transform(wavelet transform,curvelet transform,Radon transform,etc.).In the transformed domain,strong amplitude coefficients represent useful information,weak amplitude coefficients represent redundant information such as noise and artifacts.Noise interference during processing can be avoided by using a L1-norm sparse constraint on model space coefficients.Both the synthetic and field data examples verify the effectiveness after combining with compressed sensing in data processing.In field data processing,the final processing results strongly depend on the quality of the input data.Sparse transform can effectively improve the stability of the algorithm,and has broad application prospects in seismic data processing.