Study On Reconstruction And Denoising Of Two-demensional Seismic Signal Based On Sequential Generalization Of K-means

Electromagnetic exploration,seismic exploration are common means of exploration.Due to the limit of environmental,equipment conditions and economic factors,there are incomplete or non-uniform sampling phenomenon in seismic signal acquisition,at the same time the noise is inevitably involved.It will bring serious challenges to the subsequent seismic data processing.In this paper,based on sequential generalization of K-means dictionary under the compressive sensing and signal sparse representation framework,two-dimensional seismic signal reconstruction and denoising method is researched.The main research work in this paper is as follows:(1)Two-dimensional seismic signal fast reconstruction based on Sequential Generalization of K-means in dictionary.Classical seismic signal reconstruction method is limited to Nyquist sampling theorem,and the demand of sampling rate is higher,which result the large costs of seismic exploration.In addition,the transform basis function in traditional reconstruction method is single,which cannot adaptively match the specific structural features of the seismic.K-SVD denoising method can overcome the above problem,but it will takes longer while processing mass data.This paper presents a two-dimensional seismic signal fast reconstruction based on Sequential Generalization of K-means in dictionary.The basic idea is to use SGK dictionary to get much more adaptive sparse representation dictionary of seismic signal,instead of the transform basis function in compressive sensing theory;then treat the lack of seismic signals as the measurement matrix;finally the regularized orthogonal matching ROMP algorithm is adopted to rebuild the missing seismic trace.A comparison based on synthetic and real seismic experimental results shows that the proposed reconstruction method is rather rapid and better than K-SVD reconstruction method.(2)Random noise attenuation of seismic signal based on Sequential Generalization of K-means in dictionary.Seismic signal usually contains a variety of element types and structure characteristics,and a single transform basis function cannot match the following structure.K-SVD denoising method can automatically match these model structure by using adaptive transform basis functions,so the denoising effect is guaranteed.But the construction of the K-SVD dictionary takes a long time,and with the increasing of the number of iterations,the amount of data,denoising method takes longer.To this end,this paper builds a seismic signal denoising method based on a learning dictionary SGK.The basic idea is dictionary learning with noise seismic signal,and to attenuate the noise part in the seismic signal while getting the optimal SGK dictionary.The experiments show that the construction time of SGK method is far less than K-SVD method.From the experiment of the synthetic seismic signal,pre-stack marine seismic signal and post-stack land seismic signal,we verified the accuracy and rapidity of the method.(3)Random noise attenuation of seismic signal based on Sequential Generalization of K-means by controlling noise incursionIn the process of handling with seismic data containing noise signal,some noise is inevitably involved in dictionary update step using SGK denoising method.In order to control the noise incursion on dictionary atoms,this paper builds a seismic signal denoising method based on C-SGK dictionary learning under the compressive sensing framework.The basic idea is to compare the size of signal-to-noise ratio and setting threshold in dictionary update step,which will determine whether to update the atom or not.If the signal-to-noise ratio is greater than the setting threshold,sequentially update atoms,otherwise don not update.Finally the numerical experiments in this paper show that the proposed method can control the noise incursion,and also can improve the denoising effect comparing with SGK denoising method.