物探化探计算技术2025，Vol.47Issue(6)：913-922,10.DOI:10.12474/wthtjs.20241025-0005
基于残差的CBAM-UNet地震数据高分辨处理
High-resolution processing of CBAM-UNet seismic data based on residuals
摘要
Abstract
In oil and gas exploration,high-resolution seismic data is crucial.Traditional processing methods like deconvolution and inverse Q filtering are computationally complex due to their reliance on specific assumptions.Deep learning methods,which can automatically learn intricate features from data,have greatly improved the efficiency of high-resolution seismic data processing.However,the effectiveness of many current deep learning approaches in enhancing seismic resolution remains suboptimal.To address this,we have enhanced the U-Net model by integrating residual modules and Convolutional Block Attention Modules(CBAM).Adding residual modules enhances the model's ability to extract features and maintain training stability,while CBAM focuses the model on critical seismic data features,thus improving resolution.Tests on real-world seismic data show that this improved U-Net delivers superior results in high-resolution seismic data processing.关键词
高分辨率/深度学习/残差模块/注意力Key words
high resolution/deep learning/residual block/attention分类
天文与地球科学引用本文复制引用
李云涛,徐明华,邓羽淇,陈磊,文晓涛..基于残差的CBAM-UNet地震数据高分辨处理[J].物探化探计算技术,2025,47(6):913-922,10.基金项目
四川省项目(2023ZYD0158) （2023ZYD0158）
中国海洋石油有限公司"十四五"重大科技项目(KJGG2022-0903) （KJGG2022-0903）
