Seismic Data Denoising Method Based On Convolutional Neural Network

Seismic data denoising is a critical step in seismic data processing.Seismic data are required to have a high signal to noise ratio(SNR)in the subsequent data processing and analysis,but the SNR of field seismic data is affected by many factors and cannot well meet the requirements,so we need to denoise it.In recent years,deep learning technology has developed rapidly at home and abroad and has been well applied in various fields.Among them,convolutional neural networks have been applied by many researchers and have obtained important research results because of their powerful feature learning and classification capabilities.The author of this article has found through literature research: Although the neural network has been successfully used in many fields,but its application in combination with seismic data processing is still relatively small.Therefore,this paper first explores the development history and basic principles of convolutional neural networks,and the current research status of geophysical denoising methods.On this basis,an intelligent denoising method for seismic data based on convolutional neural networks is proposed.Using caffe to train the model.Then the trained model is used for seismic denoising.Experimental results show that the denoising method using convolutional neural networks is feasible.Then we discuss three important parameters of the network model: the number of convolution kernels,the size of the convolution kernel,and the number of network layers on the denoising effect.Finally,this paper using six different gradient descent methods to speed up the convergence of the network,adding regularization item to constraint the network,and using similar structural information to improve denoising quality.The experimental results show that the three improved algorithms can obtain better denoising results.