长江大学学报（自然科学版）2025，Vol.22Issue(2)：1-11,11.
基于残差收缩网络的地震数据增强方法
Seismic data enhancement method based on residual shrinkage network
摘要
Abstract
High quality seismic data is essential for seismic inversion and interpretation missions.Deep learning technology has been widely used to improve the quality of seismic data,but the existing methods are often result-oriented and have low interpretability,and the low reproducibility limits the validation of research results.Based on the characteristics of seismic data and specific processing tasks,this paper designs a unique network for seismic data enhancement to provide insights into the"black box"characteristics of neural networks in seismic exploration.The network adopts full convolution and non-downsampling framework to adapt to the high local correlation and low global correlation of seismic data.The network also combines the improved residual contraction module to improve the signal-to-noise ratio,and uses absolute maximum normalization to transform the distribution domain to make the data conform to the normal distribution with zero seismic wavelet mean.Using a strategy of training based on publicly available synthetic seismic data and testing on publicly available real seismic data,this paper verifies the effectiveness of the network in attenuation of random noise,fidelity and relative amplitude preservation through the use of signal-to-noise ratio,local similarity and spectrum analysis.The robustness of the proposed network under different noise levels and its ability to generalize real seismic data are confirmed by a number of comparative experiments with easy to reproduce and strictly controlled variables.关键词
深度学习/残差收缩网络/地震数据增强/可复现性Key words
deep learning/residual shrinkage network/seismic data enhancement/reproducibility引用本文复制引用
陈伟,陈名德,郭锐,杨浪..基于残差收缩网络的地震数据增强方法[J].长江大学学报（自然科学版）,2025,22(2):1-11,11.基金项目
非常规油气省部共建协同创新中心开放基金项目"鄂西页岩储层地震弱信号智能检测方法研究"(UOG2024-19). （UOG2024-19）
