Research On High Precision Seismic Data Reconstruction Based On Curvelet Transform

Along with the gradually increased development of seismic exploration projects in complex area,the complex terrain conditions lead to field data collection work has become increasingly difficult.Coupled with the constraints of the field exploration cost,in many cases it was very difficult for us to collect complete and conform rule-requirement seismic data.To improve the seismic exploration of resolution and SNR,the following of other processing required integrity and regular seismic data.Therefore,through the interior effective reconstruction method to solve the problems has become the focus of many scholars.This paper made a research on reconstruction of seismic data based on the theory of seismic data reconstruction,introducing the theory frame of compressed sensing,using the transform of multi-scale and multi-direction curvelet to make the sparse seismic signal,and using the method of random sampling and iterative threshold method.First,comparing three different threshold parameter formulas,selecting the best new index threshold parameter formula,then put it into the iterative threshold method.In sparse representation,in order to reflect the advantages of curvelet transform,the seismic data on the same lack using fourier reconstruction,as compared the effect of the curvelet transform and fourier transform reconstruction.About the sampling,according to the disadvantage of the random undersampling,introducing the method of one-dimensional jitter undersampling,made the comparison and analysis of the effect about the two kinds of sampling methods in reconstruction,and obtained the relation diagram between SNR and two different sampling methods of different sampling rate and the iteration times.In promoting the sparse solution,in order to show the advantages of this method,also introduced the recovery of the seismic data reconstruction by spectral gradient projection method,and compared their advantages and disadvantages by sampling rate,SNR and operation time.In order to test the method of SNR,also to add noise model for the missing trace reconstruction.At the same time,based on the 2D seismic data reconstruction,we also did the one-dimensional and two-dimensional random sampling of the 3D data time slices,showed that using this method for restoration and reconstruction have achieved good results.Finally the method was applied to the field of 2D seismic data reconstruction,and achieved good results,verified the practicability of the method.