地球物理学进展 ›› 2020, Vol. 35 ›› Issue (2): 642-655.doi: 10.6038/pg2020CC0476

• 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

深度学习在地球物理中的应用现状与前景

王昊1,2, 严加永1,2,*(), 付光明1,3, 王栩1,2,4   

  1. 1. 中国地质科学院,北京 100037
    2. 中国地质调查局中国地质科学院地球深部探测中心,北京 100037
    3. 东华理工大学地球物理与测控技术学院,南昌 330013
    4. 中国地质大学(北京)地球物理与信息技术学院,北京 100083
  • 收稿日期:2019-05-25 修回日期:2019-11-19 出版日期:2020-04-20 发布日期:2020-04-30
  • 通讯作者: 严加永 E-mail:yanjy@163.com
  • 作者简介:王昊,男,硕士研究生,研究方向为深部资源探测. E-mail: 709139830@qq.com
  • 基金资助:
    中国地质调查局地质调查项目(DD20190012);中国地质调查局地质调查项目(DD20160082);国家自然基金项目(41574133);国家自然基金项目(41630320);国家重点研发计划资助(2016YFC0600201);中国地质科学院基本科研业务费专项经费(YYWF201526)

Current status and application prospect of deep learning in geophysics

WANG Hao1,2, YAN Jia-yong1,2,*(), FU Guang-ming1,3, WANG Xu1,2,4   

  1. 1. Chinese Academy of Geological Sciences, Beijing 100037, China
    2. China Deep Exploration Center,China Geological Survey and Chinese Academy of Geological Sciences, Beijing 100037, China
    3. School of Geophysics and Measurement-control Technology, East China Institute of Technology, Nanchang 330013, China
    4. School of Geophysics and Information Technology, China University of Geosciences, Beijing 100083, China
  • Received:2019-05-25 Revised:2019-11-19 Online:2020-04-20 Published:2020-04-30
  • Contact: YAN Jia-yong E-mail:yanjy@163.com

摘要:

深度学习是机器学习的一个分支,是使用低维特征组合来表示高维特征组合,包含多重复杂结构或多个非线性变换的多个处理层对数据进行高维抽象的一种算法.深度学习的自我学习和适应能力使它在计算机视觉、语音识别、金融等领域得到了广泛应用.为分析深度学习在地球物理领域的应用前景,本文在介绍深度学习概况的基础上,结合实例分析了深度学习在地震探测定位、大地电磁数据处理、航空电磁数据解释等领域的应用现状.发现深度学习在处理海量数据在并行处理、进行模式识别特征提取、数据预测等方面都有巨大的优势.随着三维探测逐渐成为地球物理勘探主流趋势,必将产生海量数据,涉及大型计算和反演,深度学习在地球物理探测中应用将更加广泛.

关键词: 深度学习, 地球物理, 数据处理

Abstract:

Deep learning is a branch of machine learning. It is an algorithm that uses low-dimensional feature combinations to represent high-dimensional feature combinations, and multiple algorithms with multiple complex structures or multiple nonlinear transforms to perform high-dimensional abstraction on data. The self-learning and adaptability of deep learning has made it widely used in computer vision, speech recognition, finance and other fields. In order to analyze the application prospects of deep learning in geophysics, this paper analyzes the application status of deep learning in seismic exploration and localization, magnetotelluric data processing, and aviation electromagnetic data interpretation based on the introduction of deep learning. It is found that deep learning has great advantages in processing massive data in parallel processing, pattern recognition feature extraction, and data prediction. It is believed that as three-dimensional exploration gradually becomes the mainstream trend of geophysical exploration, massive data will be generated, involving large-scale calculation and inversion, and deep learning will be more widely used in geophysical exploration.

Key words: Deep learning, Geophysics, Data processing

中图分类号: