Study On Fine Seismic Processing Methods Based On Sparse Representation

Seismic data acquisition faces problems with complex surface,complicated working area condition,huge data volume and high acquisition cost,which cause irregular missing traces and large trace interval,resulting in irregular and sparse sampling in space.In addition,the random disturbance in acquisition and data processing can reduce the signal to noise ratio(SNR)of seismic record.The incomplete and low SNR data seriously affect the subsequent processing,attributes extraction and reservoir prediction.Sparse representation is the latest development in the field of information science and processing.So it has important practical significance to study its application in the reconstruction of missing data,the suppression of random noise and the processing of seismic attributes.In order to study the application of sparse representation in seismic,this paper firstly considers the seismic data reconstruction based on Fourier transform and compressive sensing.On the basis of it,the reconstruction method based on non-local algorithm and compressive sensing is studied.Non-local algorithm can utlize the similarity of non-local subblocks and regard it as the priori information of seismic data,which improves the reconstruction effect.In the random noise suppression with sparse representation,three different denoising methods are represented according to the different methods of dictionary construction,which are from easy to difficult.The denoising method based on independent component analysis(ICA)can overcome the shortcomings of traditional independent component analysis technique in the suppression of additive random noise.Wavelet domain sparse representation denoising method based on K-SVD dictionary training can be used in ground micro-seismic data with low SNR.The denoising method based on sparse K-SVD dictionary training can be applied to 3D seismic data.In the application of seismic attributes and reservoir prediction,the attribute fusion method based on independent component analysis(ICA)is studied and the influence of different fusion rules on final fusion result is considered.In view of the similarity of denoising and attribute fusion algorithm based on ICA,the integrated processing of denoising and attribute fusion is realized.In this way,it is possible to simplify the processing process without using other algorithms to preprocess the noisy data.