Seismic Data Reconstruction In Different Gather Areas Based On Sparse Representation

In recent years,with the development of society and the progress of science and technology,the contradiction between supply and demand of resources and minerals is prominent.It is an essential technical means to find resources through seismic exploration methods.Seismic data acquisition is the basis of seismic exploration.High quality seismic data can help subsequent processing and interpretation personnel to accurately determine the underground geological information.However,the original seismic data actually collected are often irregular or missing,and such data will cause uneven underground lighting and acquisition footprints,affecting the accuracy of subsequent seismic imaging.Moreover,with the increase of acquisition dimension,the storage space of computer is also increasing,and it is difficult to collect high-resolution seismic data due to the limitation of acquisition cost.The seismic data missing reconstruction method can ensure the accuracy of seismic data and reduce the cost of field acquisition.Therefore,it is necessary to propose high-precision reconstruction methods to restore missing seismic traces.In this paper,two-dimensional seismic data are arranged into three-dimensional data volume along the direction of shot point,detector and time.Firstly,50% one-dimensional random sampling is conducted on the direction of shot point,and the convex set projection algorithm based on curvelet transform is used for reconstruction.Through the simulation of the number of iterations and threshold parameters,the optimal parameters are obtained.The two-dimensional seismic data reconstruction of the time domain common shot gather is realized by reconstructing the time slice step by step.Then,the 3D data volume is transformed into the common offset-center gather to realize the reconstruction of seismic data in the common offset gather domain.By comparison,it is concluded that the reconstruction effect is better in the common offset gather domain.Although the reconstruction accuracy is high,it takes a long time to reconstruct in the time domain.In order to shorten the processing time and get better reconstruction results,this paper proposes the idea of reconstructing frequency slice directly.That is,the Fourier transform is used to convert to the frequency domain,and the frequency slice is directly reconstructed successively.By analyzing the reconstruction results of different gathers in the frequency domain,the frequency domain reconstruction significantly improves the operation time and reconstruction accuracy,and further verifies the advantages of common offset gather reconstruction.But using only one direction of information,reconstruction accuracy is insufficient,Therefore,two-dimensional random sampling of three-dimensional seismic data along the direction of shot-detection point is carried out.Adding a spatial direction information,Through successive reconstruction of time slice and frequency slice,3D seismic data reconstruction of common shot point-detector gather in time domain and frequency domain is realized.Then it is transformed into the common offset-central point domain to reconstruct the 3D seismic data of the common offset-central point gather in the time domain and frequency domain.Through comparison,it is known that in the two-dimensional random sampling,the reconstruction effect of the common offset-center point domain is slightly better than that of the common shot-detection point domain,but the operation time is long.On this basis,In order to verify the effectiveness of the proposed method,different sampling rate data and noised data are reconstructed and compared in different gather frequency domains.Finally,the optimal method is applied to the actual three-dimensional data volume,and the processing results verify the practicability of the proposed method.