物探化探计算技术2025，Vol.47Issue(6)：791-805,15.DOI:10.12474/wthtjs.20250910-0002
地震数据重建:物理模型与数据驱动的融合之路
Seismic data reconstruction:fusion of physical model and data-driven methods
摘要
Abstract
In seismic exploration,environmental and economic factors cause missing seismic data,which severely impacts subsequent imaging results.Therefore,seismic data reconstruction is a key step in obtaining high-resolution subsurface images.We investigate the development of related methods for seismic data reconstruction and give future directions.We first introduce the mathematical model of seismic data acquisition and reconstruction,then review existing methods by categorizing them into three main categories:1)model-driven methods based on physical priors;2)data-driven methods based on data feature learning;3)data and model-driven methods that fuse the two.Reconstruction methods have evolved from relying on a single physical prior to deeply integrating data features.Future seismic data reconstruction technology may focus on time-lapse seismic data reconstruction,multimodal data fusion reconstruction,and reliability analysis of reconstruction results.关键词
地震数据重建/模型驱动/数据驱动/数模双驱动/深度学习Key words
seismic data reconstruction/model-driven/data-driven/data and model-driven/deep learning分类
天文与地球科学引用本文复制引用
于四伟,陈尧,徐英杰,林荣智..地震数据重建:物理模型与数据驱动的融合之路[J].物探化探计算技术,2025,47(6):791-805,15.基金项目
地球深部探测与矿产资源勘查国家科技重大专项(2024ZD1002700) （2024ZD1002700）
国家自然科学基金项目(42574172,42574171) （42574172,42574171）
黑龙江省优秀青年基金(YQ2024D010) （YQ2024D010）
中央高校基本科研业务费(XNJKKGYDJ2024009,HIT.OCEF.2023029) （XNJKKGYDJ2024009,HIT.OCEF.2023029）
