基于迭代收缩阈值网络的地震数据重构研究  

Study on Seismic Data Reconstruction Based on Iterative Shrinkage Threshold Network

作　　者：范帅 邢磊[1,2] 李倩倩 Fan Shuai;Xing Lei;Li Qianqian(Key Lab of Submarine Geosciences and Prospecting Techniques,Ministry of Education,Ocean University of China,Qingdao Shandong 266100,China;Laboratory for Marine Mineral Resources,Qingdao National Laboratory for Marine Science and Technology,Qingdao Shandong 266071,China)

机构地区：[1]中国海洋大学海底科学与探测技术教育部重点实验室,山东青岛266100 [2]海洋国家实验室海洋矿产资源评价与探测技术功能实验室,山东青岛266071

出　　处：《工程地球物理学报》2021年第6期873-880,共8页Chinese Journal of Engineering Geophysics

基　　金：山东省重点研发计划(编号:2019GHY112019);国家自然科学基金(编号:91958206)。

摘　　要：由于复杂地球物理条件的影响,野外采集的地震数据往往存在缺失道、坏道的情况,严重影响后续的处理解释工作。压缩感知理论的提出使得地震数据可以不满足Nyquist频率进行稀疏压缩恢复,但重构效果受限于变换域以及参数的选择。因此,本文通过结合压缩感知凸优化算法迭代收缩阈值算法以及深度神经网络,通过用深度神经网络的每一层表示迭代收缩阈值算法的迭代过程,通过端到端的学习自动更新网络中的参数。将本文的方法应用于Tesseral模拟数据以及实际地震数据的地震数据重构,并与传统的迭代收缩阈值算法进行对比,实验结果表明,基于迭代收缩阈值网络的重构方法精度高,重建所需时间短,可以更有效地恢复地震信号。Due to the influence of complex geophysical conditions,the seismic data collected in the field often includes missing traces and bad traces,which seriously affects the subsequent processing and interpretation work.The theory of compressed sensing enables sparse compression recovery of seismic data that do not meet Nyquist frequency,but the reconstruction effect is limited by the selection of transform domain and parameters.Therefore,in this paper,the iterative shrinkage threshold algorithm of the compressed sensing convex optimization algorithm and the deep neural network are combined,and each layer of the deep neural network is used to represent the iterative process of the iterative shrinkage threshold algorithm,and the parameters in the network are automatically updated through end-to-end learning.The method in this paper is applied to the reconstruction of seismic data of Tesseral simulation data and actual seismic data,and compared with the traditional iterative shrinkage threshold algorithm.The experimental results show that the reconstruction method based on the iterative shrinkage threshold network has high accuracy,short reconstruction time and can recover seismic signals more effectively.

关 键 词：地震数据重构 迭代收缩阈值算法 深度神经网络 迭代收缩阈值网络 

分 类 号：P631.4[天文地球—地质矿产勘探]