Research On Random Noise Attenuation Method For Seismic Data Based On Deep Learning

In seismic exploration,the acquisition of data is interfered by many factors,and the seismic data often contains a lot of noise,which seriously affects the accuracy of seismic data processing and interpretation.To obtain high-precision seismic data processing and interpretation results,high signal-to-noise ratio data is very necessary,so suppressing seismic noise plays an important role in all aspects of seismic data processing.The traditional denoising algorithm has a good noise reduction effect,but is often limited by certain assumptions,the generalization performance is insufficient,and the calculation amount is huge,it takes a lot of time to adjust parameters,and it is difficult to deal with large data processing in practical applications.In recent years,deep learning has been widely used in the field of geophysics,and has performed well in automatic fault identification,first-arrival picking,data reconstruction,inversion,and random noise suppression.Based on the U-Net framework,this thesis proposes an improved neural network RAU-Net.RAU-Net uses the Residual Block as the basic structure of the U-shaped network to participate in the encoding and decoding process,so that the network depth is improved on the basis of the original U-Net,so as to obtain the ability to extract deeper features.In this thesis,an Attention Block is added to the RAU-Net network framework to enhance the ability of the neural network to extract effective feature information in complex backgrounds.RAU-Net adopts a supervised learning method combined with a residual learning strategy,which can better separate the effective signal and noise from the noisy profile.Compared with the original U-Net and traditional denoising methods,this algorithm has improved the performance of random noise suppression,and achieved good denoising effect.