The Suppressing Of Ground-roll In Seismic Data Based On Sparse Transform And Morphological Component Analysis

Seismic exploration is an important means to survey oil natural gas and other mineral deposits,we can infer the geological form and structure,and then find the oil and gas traps through the observation and analysis of the artificial excitation seismic wave.With the continuous development of exploration technology,the geological environment of the exploration area has become increasingly complicated and concealed,which makes the original seismic data contain more serious interference,thus affects the subsequent interpretation of seismic data and the accuracy of reservoir judgment.Therefore,it is particularly important to effectively remove the noise in the seismic data and obtain high signal-to-noise ratio,high resolution and high fidelity seismic records.Ground-roll is a kind of coherent disturbance which is more common in seismic exploration.Removing ground-roll in seismic records by appropriate methods is of great significance for the subsequent processing and interpretation of seismic data.Sparse representation is a concise and effective way to express a signal,which can represents the original signal by a linear combination of the base atoms as little as possible.Nowadays,sparse representation has been widely applied to many fields of signal processing due to its own superiority.Combing the sparse decomposition theory with seismic signal processing and making full use of the sparsity of seismic signal after decomposition in redundant dictionaries can effectively promote the process of noise suppression in seismic data processing.Based on the theory of sparse representation,this thesis studies the applications of sparse transform and morphological component analysis method in surface wave suppression.The main work and innovations are summarized as follows:(1)We firstly introduce the knowledge of sparse representation theory.Two commonly used sparse decomposition algorithms and sparse representation dictionaries are introduced in detail,and we mainly focus on the discrete cosine transform,wavelet transform and curvelet transform dictionaries.Meanwhile,the advantages and disadvantages in sparse representation performance of each kind of transform dictionary are analyzed.(2)We have researched the development status of sparse transform applied to ground-roll suppression of seismic data,and have explored two commonly used sparsetransforms-wavelet transform and curvelet transform.Based on the study of curvelet transform,the idea of secondary curvelet transform is introduced,and the direction and scale of ground-roll can be more precisely positioned.Then after the corresponding threshold processing,we can obtain the final results by curvelet inverse transform.Comparing with the results of wavelet transform and primary curvelet transform,the effect of secondary curvelet transform is better than that of these two methods,and the signal-to-noise ratio is also improved obviously.(3)We have studied the theory of morphological component analysis based on the sparsity and morphological diversity of signal,and applied it to suppress ground-roll.Seismic record has a large amount of data and complex components.Using morphological component analysis method to represent different components of seismic record sparsely by dictionaries with various atomic characteristics can characterize the main features of seismic data more effectively.On this basis,this thesis proposes discrete curvelet transform and 2D-LDCT as the sparse representation dictionaries of ground-roll and reflected wave respectively.Then,using block coordinate relaxation algorithm to solve the sparse representation model built by this two dictionaries can better realize the separation and suppression of ground-roll.Finally,we verify the validity and superiority of the method through contrast experiments.