地球物理信号特征识别与解释的机器学习方法及应用综述

Review of machine learning and application of geophysical signal feature recognition and interpretation

地球物理信号是地下介质对物理场的响应,其特征是解释地下结构和性质的主要依据.但受限于地下介质构造及物性分布特征的复杂性,地球物理信号特征的识别和解释具有不确定性.机器学习基于数据与特征的映射关系为判别地球物理信号特征和解释提供了新的思路和方法.本文围绕机器学习方法在地球物理信号特征识别及解释应用主题,梳理得到机器学习用于地球物理信号特征识别与解释的一般逻辑思路和工作流程,在提炼机器学习所涉及的处理技术和评价体系的基础上,进一步总结了机器学习在解决岩石图像识别与分类、地层岩性预测与成图、地震事件检测和到时提取、微小地震信号解释等问题时的技术要点;并对深度学习模型和简单的机器学习模型针对不同地球物理信号进行特征识别与解释的适用性和应用实效进行了分析.针对目前的发展趋势和已有研究,对机器学习在地球物理信号特征识别应用方面进行了讨论和展望.

The geophysical signal is the response of the underground medium to the physical field,and its characteristics are the main basis for explaining the underground structure and properties.However,due to the complexity of the structure and distribution characteristics of the underground medium,the identification and interpretation of the characteristics of the geophysical signal are uncertain.Based on the mapping relationship between data and features,machine learning provides new ideas and methods for distinguishing and interpreting geophysical signal features.We introduce the main content,category,and basic learning process of machine learning before proposing general logic ideas and workflows for the identification and interpretation of geophysical signal characteristics using machine learning.Then,on the basis of refining processing technology and evaluation metrics involved in machine learning,we further summarize the technical points of machine learning in solving geophysical problems such as rock image recognition and classification,formation lithology prediction and mapping,seismic event detection and phase picking,and interpretation of small seismic signals.In the meantime,ability and application effectiveness of deep learning model and simple machine learning model for feature recognition and interpretation of different geophysical signals are analyzed.Finally,we discuss and prospect the deep application of machine learning to geophysics in the future according to the current development trend and existing research.