Research On Time-variant Mixed-phase Wavelet Estimation Based On Spectral Modeling In Time-frequency Domain

Due to the stratigraphic absorption and attenuation,the real seismic record is a signal with dynamic attenuation,meanwhile the estimated wavelet varies with time.However,the traditional wavelet estimation methods based on static convolution model get a single time-invariant wavelet,which does not conform to the actual situation.In order to solve the problem that the seismic wavelet estimated by traditional methods was inaccurate because of the stratigraphic attenuation effect,the thesis analyzed the defects of the conventional segmented time-variant wavelet estimation method,and proposed the time-variant mixed-phase seismic wavelet estimation method based on spectral modeling in time-frequency domain.Compared with the segmented wavelet estimation,the method proposed here fully restored the time-variation feature of the wavelet and estimated more accurate wavelet.Considering the limitation existed in polynomial spectral modeling when fitting the wavelet amplitude spectrum,the thesis proposed the time-variant seismic wavelet estimation method based on quadratic spectral modeling in time-frequency domain,which improved the estimation precision of time-variant wavelet.At first,for the problem that the conventional segmented time-variant wavelet estimation method couldn't fully restore the the time-variation feature of the wavelet and the wavelet estimated was inaccurate,the thesis proposed the time-variant mixed-phase seismic wavelet estimation method based on spectral modeling in time-frequency domain.This method introduced the time-frequency filtering after transforming the seismic record into time-frequency domain using the improved generalized S-transformation.And then amplitude spectrum of time-variant wavelet was estimated using polynomial spectral modeling method in each moment.With the assumption that the wavelet phase was time-invariant,mixed phase was estimated by the bispectrum.Finally the time-variant mixed-phase seismic wavelet was received by combining the amplitude spectrum and phase spectrum.Compared with the conventional segmented time-variant wavelet estimation,this method fully retained the time-variation feature of wavelet.The synthetic examples and actual seismic data processing results verified that the proposed method could estimate more accurate wavelet and had an absolute advantage in seismic data processing.Then,for the problem that the use of the fitting polynomial in spectral modeling limited the shape of the wavelet amplitude spectrum and the stability of the solution is poor,the thesis further proposed the time-variant seismic wavelet estimation method based on quadratic spectral modeling in time-frequency domain.This method improved the estimation precision of the time-variant wavelet amplitude spectrum further.The synthetic examples and actual seismic data processing results showed that the method had a better effect in estimating time-variant wavelet on the basis of fully retaining the time-variation feature of wavelet.At last,the thesis contrasted the time-variant seismic wavelet estimation method based on quadratic spectral modeling in time-frequency domain with the segmented time-variant wavelet estimation method,and analyzed the advantages of the method proposed here in the time-variation feature and accuracy of wavelet estimated.The synthetic examples and actual seismic data processing results verified that the method broke through defects of the conventional segmented time-variant wavelet estimation method and fully showed the time-variation feature of wavelet.Thus,it provided a new way for estimating the time-variant wavelet accurately.