Time-varying Wavelet Extraction Method Using The Combination Of Spectral Modeling,Local Similarity And Adaptive Segmentation

Seismic wavelet suffers from scattering and absorption by the underground medium during propagation,which results in the attenuation of high-frequency components and phase distortion of the wavelet.Thus,the seismic wavelet is time-varying,and the seismogram exhibits non-stationary characteristics.Additionally,there are few mature and effective evaluation methods to discriminate the accuracy of wavelet extraction.To obtain high-resolution seismic data,the development of time-varying wavelet extraction and evaluation method is crucial.This thesis mainly finished the following innovative work.Firstly,in order to extract time-varying wavelets from non-stationary seismogram,a time-varying wavelet extraction method based on spectral modeling and adaptive segmentation in time-frequency(t-f)domain was presented.Specifically,the amplitude spectrums were extracted by spectral modeling method in t-f domain while the segmented wavelet phases were extracted by adaptive molecular decomposition method.Then,the segmented phases were extrapolated to every moment to solve the problem of matching time-varying amplitudes with segmented phases.Thus,the time-varying characteristic of wavelet during propagation could be reflected by this method,which could improve the accuracy of wavelets compared with traditional segmentation methods.Simulation and actual seismic data processing results showed the validity of the method.Secondly,considering that the fitting polynomial in spectral modeling method limits the shape of the wavelet amplitude spectrum,and the segmentation method has the assumption that the data in a segment is stationary,a t-f domain time-varying wavelet extraction method based on quadratic spectral modeling and local similarity was presented,in which amplitude spectrums were extracted by quadratic spectral modeling method in t-f domain,while the bispectrum method based on higher order cumulants was used to estimate wavelet phase preliminarily and determine the phase search range,then the phase spectrums of every point were extracted with the local similarity optimization method.Thus,the time-varying wavelet could be fitted with the amplitude spectrum at every moment.Simulation and actual seismic data processing results showed that this method could estimate the wavelets of adjacent layers more accurately with the consideration of the time-varying characteristic of the wavelet.At last,in order to solve the problem that the seismic wavelet can't be evaluated directly and reduce the sensitivity of traditional evaluation criteria to noise,an evaluation criterion based on singular value decomposition(SVD)was proposed.According to the comprehensive comparison of the common evaluation criterion's anti-noise ability,the SVD_P criterion was presented,which combines the SVD technology and the Parsimony criterion to use the primary singular values components of the deconvolution results for evaluation.With the basis of the SVD_P criterion,a seismic wavelet evaluation method was proposed,which evaluated the deconvolution results using estimated wavelets and gave a secondary evaluation of wavelet precision.Simulation and actual seismic data processing results showed the validity of the method.