摘要
地震数据的解释是地震勘探中重要的一环,常规的成果解释需要经过一系列的地震数据处理.如何做到从叠前地震数据道集中直接自动解释地层构造值得探索和研究.本文提出基于深度学习采用多标签的方式来解释叠前数据.叠前数据包含着地层结构整体的信息,对地层结构多标签标注,从叠前数据中分步的提取出各构造的信息并做出预测.这样叠前数据的解释任务就分解为对地层各个构造的预测.多标签的方式可以针对解释所关注的构造做标签,提升网络对特定构造的关注度以得到更好的预测效果.本文建立了一个从叠前数据中自动提取特征的深度学习网络,在两类数据集上建立构造标签来做预测.第一类是在Marmousi模型中随机镶嵌一个等轴状异常体;第二类地层模型含有平层、单斜、背斜和向斜等结构.检验结果表明了基于深度学习多标签的方法可以自动化的解释叠前数据预测地层构造.
Interpretation of seismic data is an important part of seismic exploration.Conventional interpretation requires a series of seismic data processing.How to interpret the stratigraphic structure directly and automatically from the pre-stack seismic data trace is worth exploring and studying.This paper proposes a multi-label supervised learning method based on deep learning to interpret pre-stack seismic data.The pre-stack data contains the overall information of the stratigraphic structure.Extract the information of each structure step by step from the pre-stack data by multi-label labeling of stratum structures.In this way,the pre-stack data interpretation task is divided into the prediction of each structure of the strata.The multi-label method can be used to label the structure of interest.It increases the network’s attention to specific structures to get better prediction results.We build a deep learning network which can automatically extracts features from pre-stack data.Two types of data sets are labeled to test our method.The first one is complex Marmousi model plugging a random inlay of an isometric anomaly;the second one contains flat,monoclinic,anticline,and syncline structures.The test results show that the multi-label method based on deep learning can automatically interpret the pre-stack data to predict the stratigraphic structure.
作者
宗志敏
何登科
孙超
ZONG ZhiMin;HE DengKe;SUN Chao(College of Geoscience and Surveying Engineering,China University of Mining and Technology(Beijing),Beijing 100083,China;State Key Laboratory of Coal Resources and Safe Mining,China University of Mining and Technology(Beijing),Beijing 100083,China;Shandong Provincial Research Institute of Coal Geology Planning and Exploration,Jinan 250104,China)
出处
《地球物理学进展》
CSCD
北大核心
2022年第3期1258-1265,共8页
Progress in Geophysics
基金
国家“111计划”(B18052)
中国矿业大学(北京)越崎杰出学者计划(2019JCA01)
山东省地勘基金项目(鲁勘字(2019)13号,(2020)19号)联合资助。
关键词
深度学习
叠前地震数据解释
多标签预测
卷积神经网络
自动化解释
Deep learning
Pre-stack seismic data interpretation
Multi-label
Convolutional neural network
Automated interpretation
