Influencing Factors Analysis And Adaptive Optimization Of Sparse Pulse Deconvolution

Deconvolution is the most effective technique to improve the vertical resolution of seismic records.By compressing seismic wavelet,the interfering strata reflection can be separated,and the potential target strata can be identified effectively.Traditional deconvolution algorithms are divided into least square deconvolution and sparse pulse deconvolution.Least square deconvolution is a linear deconvolution method which attempts to enhance the high-frequency components in the effective frequency band of seismic records,but cannot recover the frequency components outside the effective frequency band of the original seismic records.And sparse pulse deconvolution transfers the traditional deconvolution method based on compressed wavelet theory to detect the magnitude and location of reflection coefficients,breaks through the limitation of effective frequency band of seismic records,greatly improves the resolution of seismic records,and effectively reduces the number of deconvolution results by introducing different probability distribution functions of reflection coefficients as sparse constraints.Sparse pulse deconvolution is a non-linear deconvolution method.In recent years,sparsity,as a feature of signal implication,has given new development direction to seismic signal processing.In this paper,the sparse decomposition theory of signal is introduced into the sparse pulse deconvolution method,and the results are obtained by the sparse regularization inversion problem.The sparse pulse deconvolution is divided into two categories.One is sparse pulse deconvolution based on traditional Bayesian theory.Another kind of sparse pulse deconvolution is based on sparse decomposition theory.There are many factors affecting the sparse pulse deconvolution method.A small error will lead to the instability and inaccuracy of the inversion results.However,up to now,no one has systematically studied these factors.This paper divides the sparse pulse deconvolution method into two categories,and discusses several influence factors of the method based on different theories.For example,for the first kind,the paper studies the multi-solution of deconvolution,wavelet information,regularization factor and regularization constraints to discuss their influence on the inversion results.For the second kind,the paper discusses the influence of over-complete atomic libraries and sparse decomposition algorithms on the inversion results.Through some numerical simulation experiments,the factors affecting the sparse pulse deconvolution method are summarized in this paper,and some conclusions are drawn,which is convenient for future studies on sparse pulse deconvolution method.Both methods can significantly recover the high frequency components beyond the effective frequency band of seismic records.In this paper,the adaptability of two different sparse pulse deconvolution methods is discussed,and the adaptability optimization is carried out.Different methods are selected according to different situations to improve the accuracy of inversion results.