Application And Research Of Gabor Wavelet Transform In Post-stack Seismic Data Processing

Seismic data is an important source for geophysicists and geological engineers to understand underground geologic structures and to find oil and gas reservoirs.Signal processing methods are rich in content and have been widely used in seismic data processing and interpretation.Signal processing methods can help engineers analyze seismic data,extract information of the underground geological structure from the seismic data efficiently and accurately and identify geological targets that are conductive to oil and gas storage.This dissertation focuses on applications of Gabor wavelets and its transform and applications of S transform in post-stack seismic data processing and interpretation.The main content is presented as follows:?.An inverse transform of Gabor wavelet transform is presented in the dissertation.Gabor wavelet transform is widely used in seismic data analysis,but Gabor wavelet does not satisfy the wavelet admissible condition which means Gabor wavelet does not have the inverse transform that is specifically for admissible wavelets.As far as we know,no inverse of Gabor wavelet transform has been proposed yet,which restrict the application of Gabor wavelet transform to some extent.We improved the Gabor wavelet,and on this basis,we proposed an inverse transform of Gabor wavelet transform which perfects theory and provides theoretical support for expanding application range of Gabor wavelet.Based on the Gabor wavelet transform and its inverse transform,a new method to improve the resolution of post-stack seismic data is presented in this dissertation.An example demonstrates that the seismic data processed by the new method has higher resolution and fidelity.?.To study S-transform and the various generalized S-transform,an ordinary type of generalized S-transform is presented in the dissertation.This type of S-transform has two important parameters,dominant frequency and resolution factor.The dominant frequency of the transform is center frequency of time-frequency windows,and resolution factor controls the resolution of the transform which is a function of the dominant frequency of the transform.In the application,the resolution factor function can select as needed.The simulation experiment demonstrates that the ordinary type of the generalized S-transform can flexibly control the resolution of the transform in time-frequency analysis of signal,and has more accurate time-frequency locating ability.?.After studying filtering methods of seismic data,we recognize that the consistency between the processing method and the physical model of the processing object is a prerequisite for getting a good result.Seismic trace is the basic unit of seismic data,and it is a one dimension time series which shows the vertical variation of strata.Taking seismic section as two-dimensional image,the filtering strategy is not completely suitable for the physical structure of seismic section.Two-dimensional smooth filters can filter out noises of seismic data,meanwhile,smooth out some noise-like information and edge information that reflects strata lateral unconformity.Edge preserving filtering can preserve some edge information,but may also strengthen the fake edge information.The filtering method that is performed on every individual seismic trace is consistent with the physical structure of seismic data.It filters out the noise of the current trace without affecting the neighboring seismic traces,which offers one possibility to preserve the lateral changes of seismic data.Hence,on the premise of filtering on each seismic trace,how to preserve edge of seismic data is worthy of further study.Finally,in order to implement this goal,we made some explorations,and proposed three one-dimensional time-domain filter which are a Gabor wavelet filter,a Gabor wavelet integral filter and a Gabor wavelet time-frequency filter respectively.The case demonstrates that the Gabor wavelet integral filter has good performance of edge preserving.?.We improved methods of attributes extraction.Firstly,notice that extracting instantaneous seismic attributes needs apply the Hilbert transform to seismic data.However,the Hilbert transform is sensitive to noise,then when it is used to calculate instantaneous seismic attributes in low SNR data,the error is relatively larger.Hence,we proposed that use a Gabor wavelet integral filter instead of the Hilbert transform to improve the quality of instantaneous seismic attributes.Secondly,the coherence attribute is a good indicator of faults,and the third-generation coherence algorithm has a stable performance but it is computationally expensive because of the eigenvalue calculation of covariance matrix.Based on the hypothesis of strata continuity,an iterative initial value selection method is proposed which effectively improves the convergence rate.Thirdly,in the dissertation,extracting method of spectral decomposition and frequency-dividing attributes are studied.Because Gabor wavelet transform is a liner-zero-phase filter and its resolution factor is self-adaptive,by comparing various time-frequency analysis methods,the Gabor wavelet transform has accurate time-frequency locating ability and flexible bandwidth control characteristics.It is proved by theoretical study that dominant-frequency harmonic of sub-band signal extracted by Gabor wavelet transform is the same as this frequency harmonic of original signal,so Gabor wavelet transform is a spectral decomposition and frequency division tool with fidelity.Lastly,we studied the application of seismic attributes in seismic data interpretation and proposed a new method of using the frequency-dividing instantaneous phase to identify small faults.A simulation and a case prove the correctness and effectiveness of the new method.?.Methods for improving the resolution of post-stack seismic data were studied systematically.Two new methods for improving the resolution based on time-frequency spectrum compensation are presented.Firstly,this dissertation discussed causes of the low resolution of seismic signals and the possibilities of improving the resolution by mathematical and physical methods.Then,utilizing the relationships among the time-frequency spectrums,and the frequency spectrum of seismic signal and the strata's attenuation based on the convolution model of seismic traces,the dissertation proposed two methods that improves resolution of seismic signal by compensating time-frequency spectrum of S transform and Gabor wavelet transform respectively.Lastly,a validation of the new method is performed.As we expected,the results sufficiently showed the fidelity and rationality of the new method,and also the results are better than traditional methods.