Reconstruction Of Multidimensional Seismic Data Based On Parallel Matrix Factorization Algorithm

In the process of ocean OBS data acquisition,due to the limitation of acquisition cost,the number of receivers in the observation system is small,the number of shots is relatively large,the uniformity of coverage times is poor,and even zero coverage occurs;OBS is affected by ocean currents during the process of sinking.The spacing of OBS is uncontrollable,resulting in irregular spacing;During the traveling process,the ship is affected by wind and waves,and the bending of the gun lines is crossed;During the course of the ship's travel,it is affected by the wind and waves,and the bending of the gun line is crossed.The above reasons cause the collected OBS data to be irregular(or sparsely distributed)in the spatial direction.In the process of collecting data on land,due to terrain conditions(mountain valleys,lakes and rivers,ore bodies and forbidden mining areas,etc.)and ground obstacles(bridge roads and buildings,etc.),the data obtained will appear irregular.Irregular seismic data have a great influence on velocity analysis,superposition and migration,resulting in unclear local velocity and poor imaging,which can not achieve the purpose of fine mapping of underground geological structures.Therefore,interpolation reconstruction of seismic data is very necessary.The Parallel Matrix Factorization(PMF)algorithm used in this paper supplements and extends the seismic data tensor reconstruction based on high-order SVD factorization and nuclear norm minimization.It can simultaneously reconstruct and denoise seismic data,but the computational efficiency increases exponentially with increasing dimensions.It is a frequency stripe in a multi-dimensional space that expands into a strip matrix according to each dimension,and then performs a rank reduction decomposition for each strip matrix,and finally folds it into a tensor and weights the sum.On the basis of the Matlab version of the PMF program,in order to improve the efficiency of the program,it is rewritten as a Fortran version of the program.Aiming at the data denoising problem,the robust PMF algorithm is introduced,which can achieve the effect of data reconstruction and random noise removal in the conventional PMF algorithm,and can suppress abnormal noise.In this paper,the robust PMF algorithm is used to interpolate and reconstruct 4D model data,ocean OBS 4D data and 5D land data.This method is not only suitable for high-resolution,high signal-to-noise ratio ocean data,but also has a good effect in processing land data.The results show that the parallel matrix factorization algorithm can effectively solve the problem of irregular and sparse distribution of seismic data,and the application effect is good.