Research On Fast Matching Pursuit And Reconstruction For Missing Seismic Trace

At the area of seismic exploration,the research on sparse decomposition of seismic trace is in the ascendant.The precision and speed of decomposition have a far influence on post processing effects of seismic data.Matching pursuit has become more and more widely used in the seismic data processing area because it can represent seismic trace linearly according to the characteristic of time-frequency.However,the amount of calculation is so large that the data processing becomes inefficient.As for this problem,this paper proposed a fast matching pursuit method based on genetic algorithm and orthogonal atom.Genetic algorithm could narrow the search range of atom dictionary and reduce the number of greedy iteration.The redundant components could be eliminated by the orthogonalization of atoms and the process of residual convergence is accelerated effectively.In order to increase the flexibility of decomposition,this paper used the adjacent residual ratio threshold as the terminating condition of iteration.The proposed method is applied to sparse decomposition of synthetic seismic trace and real seismic trace respectively,and the experimental results show that this method not only could reduce the sparsity of decomposition but could improve the operating speed greatly.Secondly,the seismic trace often appears incomplete or irregular because of the influence of environment and cost in real seismic exploration.As for this problem,this thesis presents a reconstruction method based on compressed sensing theory model and K-SVD dictionary learning.At first,the initial DCT dictionary is trained by lots of seismic data samples in order to acquire a overcomplete dictionary matching with the characristics of reconstructed data.This new overcomplete dictionary replaces the traditional fixed function.Then the sampling matrix of missing seismic data is used as the measurement matrix of compressed sensing model.Staged regularized orthogonal matching pursuit is applied to the reconstruction phase of compressed sensing.Through the experiment of synthetic seismic trace and real seismic data,we find that the reconstruction method of this thesis has great effect on missing seismic traces and verifies the validity and feasibility of the algorithm.At last,this thesis presents a new method of time-frequency analysis for seismic trace by combining the basis pursuit algorithm and Wigner-Ville distribution.The seismic trace is decomposed into an optimal superposition of a series of time-frequency atoms based on basis pursuit.Then the Wigner-Ville distribution of every atom is added to reflect the time-frequency features.The method of time-frequency analysis based on Basis Pursuit not only reduces the crossing items caused by Wigner-Ville distribution but improves the time-frequency resolution.