一种基于DNDCNN的地震信号去噪方法
DOI:
作者:
作者单位:

太原理工大学

作者简介:

通讯作者:

中图分类号:

TP391

基金项目:

中央引导地方科技发展基金(No.YDZJSX2021C004,No.YDZJSX20231C004);山西省自然科学基金(No.20210302124554)


A seismic signal denoising method based on DNDCNN
Author:
Affiliation:

Taiyuan University of Technology

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    在复杂的勘探环境下,原始采集的地震数据包含大量随机噪声会严重影响地震资料的质量,为后续地质解释带来困难。针对该问题,本文提出了一种结合可变形卷积和注意力机制的地震信号去噪模型,即DnDCNN(Denoising Deformable Convolutional Neural Network)网络。首先,在DnCNN网络中引入融合可变形卷积的注意力机制,使网络更加关注有效信号区域,减少细节信息的丢失;其次,将网络中堆叠的标准卷积替换为可变形卷积和标准卷积串联模式,提高不变性特征提取能力;最后,将批量归一化和残差学习策略融合,实现网络快速收敛和信噪分离。通过对模拟和实际地震数据进行验证,结果表明该网络模型在不同噪声水平下可以有效压制随机噪声、保留更多细节信息,对微弱信号去噪表现出更优秀的信噪比。

    Abstract:

    In complex exploration environments, the raw seismic data collected contains a large amount of random noise, which seriously affects the quality of seismic data and brings difficulties for subsequent geological interpretation. In response to this issue, this article proposes a seismic signal denoising model that combines attention mechanism and deformable convolution, namely the DnDCNN (Denoising Deformable Convolutional Neural Network) network. Firstly, the attention mechanism integrating deformable convolution is introduced into the DnCNN network to make the network pay more attention to effective signal regions and reduce the loss of detailed information; Secondly, replace the stacked standard convolutions in the network with deformable convolutions and standard convolution concatenation modes to improve the ability of invariant feature extraction; finally, the batch normalization and residual learning strategies are integrated to achieve rapid network convergence and signal-to-noise separation. Through verification of simulated and actual seismic data, the results show that the network model can effectively suppress random noise, retain more detailed information, and perform better signal-to-noise ratio for weak signal denoising under different noise levels.

    参考文献
    相似文献
    引证文献
引用本文

马俊卓,李钢,孙嘉莹,等. 一种基于DNDCNN的地震信号去噪方法[J]. 科学技术与工程, , ():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2023-10-31
  • 最后修改日期:2024-05-21
  • 录用日期:2024-05-21
  • 在线发布日期:
  • 出版日期:
×
亟待确认版面费归属稿件,敬请作者关注