文献类型：期刊文献

英文题名：Attenuation of seismic swell noise using convolutional neural networks in frequency domain and transfer learning

作者：You, Jiachun[1];Xue, Yajuan[2];Cao, Junxing[1];Li, Canping[3]

机构：[1]Chengdu Univ Technol, Sch Geophys, Chengdu 610059, Sichuan, Peoples R China;[2]Chengdu Univ Informat Technol, Sch Commun Engn, Chengdu 610225, Peoples R China;[3]Guangdong Ocean Univ, Lab Ocean Remote Sensing & Informat Technol, Zhanjiang 524088, Peoples R China

年份：2020

卷号：8

期号：4

起止页码：T941

外文期刊名：INTERPRETATION-A JOURNAL OF SUBSURFACE CHARACTERIZATION

收录：SCI-EXPANDED(收录号：WOS:000606163500050)、、Scopus(收录号：2-s2.0-85089732000)、WOS

基金：The authors wish to thank the editors K. Marfurt and X. Wu and express their gratitude to the anonymous reviewers and N. Pan for their insightful and constructive comments that greatly improved this paper. This work was supported by the National Natural Science Foundation of China (grant no. 41430323) and Innovation and Strong School Foundation of Guangdong Ocean University (grant no. 230419096). Special thanks also go to the SINOPEC Key Laboratory of Geophysics for its support.

语种：英文

外文摘要：Because swell noises are very common in marine seismic data, it is extremely important to attenuate them to improve the signal-to-noise ratio (S/N). Compared to process noises in the time domain, we have built a frequency-domain convolutional neural network (CNN) based on the short-time Fourier transform to address swell noises. In the numerical experiments, we quantitatively evaluate the denoising performances of the time- and frequency-domain CNNs, compare the impacts of network structures on attenuating swell noises, and study how network parameter choices impact the quality of the denoised signal based on peak S/N, structural similarity, and root-mean-square-error indices. These results help us to build an optimal CNN model. Furthermore, to illustrate the superiority of our proposed method, we compare the conventional and proposed CNN methods. To address the generalization capability of CNN, we adopt transfer learning by using fine tuning to adjust the weights of the pretrained model with a small amount of target data. The application of transfer learning improves the quality of the denoised images, which further proves that our proposed method with transfer learning has the potential to be deployed in actual seismic data acquisition.