地球物理学进展 ›› 2020, Vol. 35 ›› Issue (6): 2211-2219.doi: 10.6038/pg2020DD0494

• 应用地球物理学Ⅰ(油气及金属矿产地球物理勘探) • 上一篇    下一篇

地震数据智能去噪与传统去噪方法的对比及展望

孙小东1,2(), 王伟奇1, 任丽娟3, 李振春1   

  1. 1.深层油气重点实验室,中国石油大学(华东),青岛 266580
    2.山东省油藏地质重点实验室,中国石油大学(华东),青岛 266580
    3.中海石油(中国)有限公司湛江分公司,湛江 524000
  • 收稿日期:2019-12-11 修回日期:2020-08-17 出版日期:2020-12-20 发布日期:2020-12-26
  • 作者简介:孙小东,男,1980年生,博士,中国石油大学(华东)地学院地球物理系讲师,研究领域为地震波偏移成像和反演.E-mail: sunxd@upc.edu.cn
  • 基金资助:
    国家自然科学基金项目(41574098);国家自然科学基金项目(41630964);中央高校基本科研业务费专项资金(18CX02059A);深层油气重点实验室开发基金(20CX02111A);中国石化地球物理重点实验室开放基金项目(wtyjy-wx2018-01-07);中石油重大科技项目(ZD2019-183-003)

Comparison and prospect on AI denoising of seismic data along with traditional denoising methods

SUN Xiao-dong1,2(), WANG Wei-qi1, REN Li-juan3, LI Zhen-chun1   

  1. 1. Key Laboratory of Deep Oil and Gas, China University of Petroleum (East China), Qingdao 266580, China
    2. Shandong Provincial Key Laboratory of Reservoir Geology, China University of Petroleum (East China), Qingdao 266580, China
    3. Zhanjiang Branch of CNOOC (China) Company Limited, Zhanjiang 524000, China
  • Received:2019-12-11 Revised:2020-08-17 Online:2020-12-20 Published:2020-12-26

摘要:

在油气地震勘探中,提高地震资料的分辨率、信噪比、保真度一直以来都是地震资料处理环节的关键问题.一般来说,高信噪比是高分辨率的前提,只有信噪比达到一定的水平,地震资料处理才可能实现良好的保真度.当今的地震勘探正向着深层、复杂构造和岩性勘探的方向发展迈进,人们相应的对地震资料处理也提出了更高的要求.目前,业界比较流行的去噪手段包括f-x RNA法、f-x-y NRNA法、3D t-x-y APF法和t-x SOPF法等.本文将这些传统去噪方法与压缩感知去噪、卷积神经网络法等先进的智能去噪技术做了对比,并对智能去噪技术的前景做进一步的展望.

关键词: 去噪, 滤波器, 正则化, 卷积神经网络

Abstract:

In oil and gas seismic exploration, improving the resolution, signal to noise ratio and fidelity of seismic data is always the key problem in seismic data processing. In a general way, high SNR is the precondition of high resolution. Only when SNR reaches a certain level, seismic data processing can achieve good fidelity. Nowadays, seismic exploration is developing towards the direction of deep, complex structure and lithology exploration, so people put forward higher requirements for seismic data processing. Currently, popular denoising methods in the industry include f-x RNA, f-x-y NRNA, 3D t-x-y APF and t-x SOPF. This paper compares these traditional denoising methods with advanced intelligent denoising technologies such as compressed sensing and convolutional neural network.

Key words: Denoising, Filter, Regularization, Convolutional neural network

中图分类号: