Research On Denoising Method Of DAS Data Based On Sparse Representation

With the development of oil and gas exploration technology and the environmental factors of seismic data acquisition in the field,there are serious noise pollution problems in seismic data.Noise pollution will reduce the signal-to-noise ratio of seismic data,increase the difficulty of seismic data processing technology and affect subsequent seismic data interpretation.In order to ensure the quality of seismic data,it has become an urgent requirement in the current related fields to study the methods of noise reduction and processing of the original seismic data collected in the field.Firstly,in this paper,we introduces the necessity of noise reduction in seismic exploration,signal acquisition and other practical applications.Secondly,we explains its important position in seismic exploration.Then we introduce the research status of noise reduction processing methods at home and abroad.Finally we introduces the basic principles related to the noise reduction methods,such as sparse representation,tensor theory and so on.According to the characteristics of seismic signal and noise,two kinds of noise reduction methods are proposed.Specifically,it includes the following research contents:(1)Aiming at the problem that there are a lot of regular broken-line noise and random Gaussian noise in distributed fiber acoustic sensor data,this paper combines morphological component analysis with low-rank sparse representation theory,and proposes a DAS data denoising method based on low-rank sparse representation.In this method,morphological component analysis is introduced to treat noise and effective signal as two components.According to the characteristics of signal,a suitable dictionary is found.Then a low-rank sparse representation model combining morphological component analysis is established and solved by using low-rank and sparsity.The effective signal and noise are separated.Finally,the effective signal is reconstructed by low-rank sparse expression matric,which has good performance.(2)Aiming at the problem that traditional sparse coding denoising methods are usually used in two-dimensional seismic data and can not use the structural information of three-dimensional seismic data to further reduce noise interference,this paper proposes a denoising method for DAS data based on weighted tensor sparse coding.We introduce weighted tensors to describe the residual and regularization items in the framework of tensor sparse coding.A weighted tensor sparse coding model is proposed based on the statistical characteristics and prior probability of DAS data.Then we transform the model into linear equality constraints by further optimization,and solve it by alternating direction multiplier algorithm to achieve the purpose of noise reduction.In this paper,two methods based on sparse representation are applied to the threedimensional DAS data of the actual seismic work area.By comparing the effect before and after noise reduction and the effect of other methods,it is found that the effect of noise reduction has reached the expectation.