Study On The Elimination And Migration Of Surface Related Multiple By Sparse-constrain Inversion

Abundant of surface-related multiple exit in the marine seismic data because of the strong reflection of marine surface.Surface multiple has been regarded as noise in the traditional seismic exploration,and its elimination is a key issue which is difficult to be solved effectively.At present,the common seismic processing and interpretation methods used in production are proposed based on primary,if the multiple of seismic data is not removed or not removed completely,it will bring interference to the post-processing,such as the multiple crosstalk in the tracked section and migration section which mislead the geological explanation.Therefore,the prediction,matching and elimination of multiple have been meaningful research topics.On the other hand,as the deepening understanding of seismic wave-field information,more and more scholars put forward that multiple information no longer only can be regarded as a kind of interference,it is the same seismic waves spread through underground like primary,carrying the information of underground structure.If processed correctly multiple can improve seismic illumination.Multiple spread more times in underground than primary,so it has bigger reflected angles,and observation aperture is bigger;On the other hand,multiple has more travel path in underground,so it can have better lighting for the steep dip angle structure and under salt,which improves the illuminating scope of seismic observation.Therefore,carring out the related research of using the multiple and promoting the use of full seismic data are the main trends of future development of multiple.This paper first reviews the classical surface-related multiple elimination method,analyzes its core algorithm and extends it to three dimensional and realizes the 3D SRME.In addition,we use the sparse reconstruction algorithm instead of the cross-line multiple contribution calculating process of traditional 3D SRME.This algorithm picks up every event of Fresnel zone information based on the available seismic cross-line data,realizes the reconstruction of multiple by the location and amplitude information of Fresnel zone.Three dimensional multiple prediction by sparse inversion(3D MPSI)can be well adapted to the 3D Marine seismic streamer data,solves the seismic data sparse problem effectively.Traditional SRME achieves the multiple removing by adaptive matching subtracting,this method is based on the least square principle,so there often will be multiple over-matching problem.In order to overcome this problem,we propose the Curvelet domain multiple match method based on the seismic data sparse representation,the Curvelet transformation has the characteristics of multi-scale and multi-direction,achieves the better matching to suppress multiple.In addition,we adopt the cubic B-spline function to improve the traditional Curvelet multiple match method,select a little number of unknowns as the basis points of the matching coefficient,apply the cubic B-spline function on these basis points to reconstruct the matching array,build constraint solving equation based on the relationship of predicted multiple,matching coefficients and actual data,use the BFGS algorithm to iteration and realize the fast solving sparse constraint of multiple matching algorithm.In order to avoid the traditional sense of multiple matching subtractive process,based on the study of Berkhout seismic data matrix,we carry out two kinds of multiple processing methods,estimate primary by sparse inversion(EPSI)and primary and multiple estimated in inverse data space.Take the L1 norm constraint to replace the L0 norm constraint of traditional EPSI,implement the primary impulse response and source wavelet updated by iteration,which omits the matching subtractive process of traditional SRME method,avoids the introduction of the interference of artificial and promots the accuracy of estimate primary to some extent.To the traditional inverse data space method,we apply 3D Curvelet sparse transform and SPGL1 sparse solution,build an inversion problem according to the relationship of seismic data,surface related operator and primary impulse response,use the L1 norm constraint to transform the inverse problem into a convex optimization problem,use SPGL1(L1)spectrum gradient projection algorithm to solve the inverse problem.Through the salt dome model data test,it is found that this method can effectively identify the surface multiples,and the energy of calculate primary is more concentrated in phase axis,and have higher resolution.In the branch of utilizing of multiple.First,we introduce the multiple illumination into underground complex structures by double acoustic seismic illumination which is based on the finite difference time domain method,and implement the illuminate analysis of Pluto complex geological structure model using primary and multiple wave-field.Compared with the traditional illumination method,the illuminated wave-field which contains multiple is richer,the salt structure is clearer,and the illumination shadow areas decrease significantly.Second,by taking research of multiple reverse-time migration,analyzing the causes of multiple cross-talk,and imitating the imaging principle of the traditional reverse-time migration,we implement multiple divided-order reverse-time migration,which can make use of multiple wave information,and avoid generating crosstalk caused by multiple.Finally,we carry out a full wave field seismic data migration method which takes the model perturbation as migration imaging results according to the born approximation.The proposed method takes multiple into the demigration operator,and builds the constraints inversion formula.And then solve the fast sparse inversion method by SPGL1 to realize the full wave field migration imaging.Compared with traditional reverse-time migration method,we can prove the superiority of our method.Through the above discussion,the processing method of multiple from the beginning to suppress as interference information,gradually develop taking advantage of the multiple information,this paper expounds the understanding a development process of multiple.Multiple prediction and separation provides the technical basis of multiple division order migration imaging,the multiple is predicted more accurate,the higher quality of migration imaging.Finally,by improving the migration imaging concept,build sparse inversion of full wave field migration imaging method,realize the full wave field contains surface related multiple wave migration method.Many methods described here is based on sparse transformation,sparse constraint solving process,introduce these methods in seismic data processing greatly improve the precision of data.Through the model data and real data verification,the multiple prediction,matching subtracting,and migration imaging have reached a certain application effect.