Node Seismic Data Recovery And Reconstruction Based On Compressive Sensing Theory

The surface acquisition environment of real reservoirs is becoming more and more complicated,causing the traditional cable instrument cannot reach the accuracy requirement for field seismic acquisition.On the contrary,the node seismic acquisition system is flexibility and simple to operation,so that it can be used to complex surface environment more efficiently.However,in the process of actual exploration projects,the nodes are easily lost,or even continuously lost,for its wireless,which usually results in the absence of seismic data in acquisition data.In some serious cases,re-acquisitions are necessary,leading to increasing workload and economic costs of field acquisition.Therefore,in view of the incompleteness of node seismic data,this paper uses compressed sensing theory to restore and reconstruct the whole data,expecting to reduce the field re-acquisition workload and save exploration cost.The core idea of compressed sensing theory is to transform the cross-coherent alias caused by traditional regular under-sampling into low-amplitude incoherent noise by using random under-sampling method,so that the real spectrum can be detected easily.Finally,the real spectrum can be restored to the original signal by some reconstruction algorithm.Its advantage is that the theory can reconstruct the original data by the sampling points far bellower than that required by Nyquist sampling theorem.Therefore,this algorithm is of great significance for the recovery and reconstruction of seismic data,which is also the reason why the compressed sensing theory is used here.The implementation of compressed sensing theory includes three steps: sparse transformation of signals,selection of sampling matrices and reconstruction of data.By analyzing the sparse transformation methods such as Fourier transform,short-time Fourier transform,wavelet transform,ridgelet transform and curvelet transform,the curvelet transform has the characteristics of multi-scale,localization and directionality,which can carry out the optimal nonlinear approximation to seismic wave front information.The designing method of sampling matrix has also developed from regular sampling to random sampling.Through the comparison of minimum norm method,matching pursuit series algorithm,minimum total variation method and iteration threshold method,it is concluded that the iteration threshold method is easy to operate,fast to iterate,and can effectively reconstruct the actual noisy data.Therefore,the compressed sensing algorithm based on curvelet transform,random sampling and iterative threshold method is selected to recover and reconstruct the node seismic data with random distribution characteristics of missing traces.The test results of model data show that when the data on both sides of missing tracks are complete and the percentage of missing tracks is less than 6%,the recovered and reconstructed data is great.In this process,the choice of threshold and scale of curvelet transform is to judge synthetically according to actual data.The same method is used to the actual node seismic data of a given 1378-shots.Compared with the original measured data,the recovered data has more continuous events and more abundant information,which indicates the validity of the compressed sensing theory for this data.