Seismic Wavelet Blind Estimation And Seismic Sparse Deconvolution

With the growing demand of oil and gas,and the situation that complicated and deep reservoirs have become the main focus of seismic exploration,the industry has put forward higher request on seismic signal processing.The deconvolution of seismic traces is one important processing method,of which seismic wavelet extraction is usually the first step.The fourth-order cumulant matching method has been proposed for estimating a mixed-phase wavelet from a convolutional process since several years ago.The characteristic of this method is that it can extract seismic wavelet without using well-logging information,under the constraint that the reflectivity is a non-Gaussian,stationary and statistically independent random process.This leads to a highly nonlinear problem,so heuristic algorithm is suitable to be applied here.I have studied a novel class of method for the global optimization of continuous variables based on simulated annealing?SA?,it's called coupled simulated annealing.The coupled SA?CSA?class is characterized by a set of parallel SA processes coupled by their acceptance probabilities.CSA-M is the most powerful one among this new class of method,as it has an online regulating mechanism that can reduce the sensitivity of the algorithm to initialization parameters,effectively avoiding the burden of parameter tuning.Meanwhile,the CSA algorithms can be executed using the technology of multi-thread,without increasing computational expense even with the existence of multiple optimization processes.Numerical tests demonstrate that CSA-M can successfully solve the fourth-order cumulant matching problem to estimate seismic wavelet,even with noise-corrupted data.Inspired by the well-developed L_1/2 regularization in sparse recovery theory,a novel method is proposed which uses L_1/2 regularization as the sparse constraint of reflection coefficient for seismic deconvolution processing.I use a specific threshold iterative algorithm to solve the L_1/2 problem.Tests on stationary seismic data demonstrate that this method can invert the reflectivity coefficients effectively,with satisfactory robustness to regularization parameter and noise.Furtherly,I extend this method to process nonstationary seismic data,synthetic examples verify that it can remove the effects of the earth's Q filtering and the wavelet's filtering.Finally,using L_1/2 regularization and block coordinate descent method,I develop a novel method that can simultaneously retrieve seismic reflectivity and seismic wavelet when initial input seismic wavelet is inaccurate,which is common in realistic scenarios.Numerical examples demonstrate its effectiveness.