Seismic velocity inversion method based on feature enhancement U-Net
ZHANG Yan1,2, MENG Decong1, SONG Liwei3, DONG Hongli2
1. School of Computer and Information Technology, Northeast Petroleum University, Daqing, Heilongjiang 163318, China; 2. Artificial Intelligence Energy Research Institute, Northeast Petroleum University, Daqing, Heilongjiang 163318, China; 3. School of Physics and Electronic Engineering, Northeast Petroleum University, Daqing, Heilongjiang 163318, China
Abstract:The challenge faced by seismic velocity inversion methods based on deep neural networks is that the weak semantic mapping correspondence between seismic data in the time domain and model information in the spatial domain leads to a high degree of multiplicity.Additionally, neural networks lack effective guidance in mapping seismic data to velocity models, making them susceptible to noise interference and thus affecting inversion accuracy.Therefore, a seismic velocity inversion method based on feature enhancement U-Net is proposed.Firstly, by integrating the features of multi-shot seismic data, the spatial relationship between the seismic time series signal input to the network and the corresponding velocity model becomes more apparent.Subsequently, based on the concept of multi-scale feature fusion, modules with convolutional kernels of varying sizes are designed to bolster the network’s capacity for learning effective features.Next, attention gates are used to guide the network and enhance the features that the network focuses on.Finally, based on the bottleneck residual and pre-activation, a pre-activation bottleneck residual is incorporated into the network, to avoid gradient disappearance and network degradation.The experiment shows that this method has higher accuracy in seismic velocity inversion and performs well in noise testing.It has a certain generalization ability.
GAROTTA R, MICHON D.Continuous analysis of the velocity function and of the normal-moveout corrections[J]. Geophysical Prospecting, 1967, 15(4):584-597.
[2]
张明, 薛诗桂.基于物理模型的起伏地表构造成像叠加速度分析研究[J].地球物理学进展, 2013, 28(3):1417-1424.ZHANG Ming, XUE Shigui.Stacking velocity analysis of structure imaging with relief surface based on seismic physical model[J].Progress in Geophysics, 2013, 28(3):1417-1424.
[3]
徐文君, 殷俊锋,王华忠,等.基于稀疏反演理论的自动叠加速度反演方法[J].地球物理学报,2017,60(7):2791-2800.XU Wenjun, YIN Junfeng, WANG Huazhong, et al.Automatic estimation of stacking velocity based on sparse inversion[J].Chinese Journal of Geophysics, 2017,60(7):2791-2800.
[4]
王瑞林, 冯波, 吴成梁, 等.基于CMP道集智能化的初始速度建模方法研究[J].石油物探, 2021, 60(5):763-772.WANG Ruilin, FENG Bo, WU Chengliang, et al.Intelligent initial velocity model building based on CMP gathers[J].Geophysical Prospecting for Petroleum, 2021, 60(5):763-772.
[5]
ROSS H N.Prestack Gaussian-beam depth migration[J].Geophysics, 2001, 66(4):1240-1250.
[6]
胡治权, 曹俊兴, 姚维益.基于射线参数估计的快速2D Kirchhoff叠前深度偏移[J].物探化探计算技术, 2008, 30(5):368-373, 348.HU Zhiquan, CAO Junxing, YAO Weiyi.High-efficiency 2D Kirchhoff pre-stack migration based on the ray-parameter estimation[J].Computing Techniques for Geophysical and Geochemical Exploration, 2008, 30(5):368-373, 348.
[7]
杜启振, 李芳, 侯波, 等.角度域弹性波Kirchhoff叠前深度偏移速度分析方法[J].地球物理学报, 2011, 54(5):1327-1339.DU Qizhen, LI Fang, HOU Bo, et al.Angle-domain migration velocity analysis based on elastic-wave Kirchhoff prestack depth migration[J].Chinese Journal of Geophysics, 2011, 54(5):1327-1339.
[8]
ZHANG K, LI Z C, ZENG T S, et al.Residual curvature migration velocity analysis for angle domain common imaging gathers[J].Applied Geophysics, 2010, 7(1):49-56, 99.
[9]
秦宁, 王延光, 杨晓东, 等.基于角道集剩余曲率分析的层析速度建模[J]. 石油地球物理勘探, 2015, 50(1):61-66.QIN Ning, WANG Yanguang, Yang Xiaodong, et al.Tomography velocity model building based on ADCIG's residual curvature[J].Oil Geophysical Prospecting, 2015, 50(1):61-66.
[10]
李勇德, 董良国, 刘玉柱.一种新的预条件伴随状态法初至波走时层析[J].地球物理学报,2017, 60(10):3934-3941.LI Yongde, DONG Liangguo, LIU Yuzhu.First-arrival traveltime tomography based on a new preconditioned adjoint-state method[J].Chinese Journal of Geophysics, 2017, 60(10):3934-3941.
[11]
沈天晶, 胡叶正, 黄旭日, 等.共聚焦点层析成像方法[J].石油地球物理勘探, 2021, 56(5):1074-1085.SHEN Tianjing, Hu Yezheng, HUANG Xuri, et al.Research on tomography method based on common focus point[J].Oil Geophysical Prospecting,2021,56(5):1074-1085.
[12]
DAHLEN F A,HUNG S H,NOLET G.Fréchet kernels for finite-frequency traveltimes-I.Theory[J].Geophysical Journal International,2000,141(1):157-174.
[13]
XIE X B, YANG H.The finite-frequency sensitivity kernel for migration residual moveout and its applications in migration velocity analysis[J].Geophysics, 2008, 73(6):S241-S249.
[14]
陈云鹏.海南地幔柱P波有限频层析成像研究[D].黑龙江哈尔滨:哈尔滨工业大学, 2020.CHEN Yunpeng.Hainan Mantle Plume from P Wave Finite Frequency Tomography[D].Harbin Institute of Technology, Harbin, Heilongjiang, 2020.
[15]
张天泽, 韩立国, 张盼, 等.基于自适应步长OBFGS算法的快速时间域全波形反演[J].物探化探计算技术, 2018, 40(1):1-7.ZHANG Tianze, HAN Liguo, ZHANG Pan, et al.Fast time domain full waveform inversion based on self-adaptive step OBFGS algorithm[J].Computing Techniques for Geophysical and Geochemical Exploration, 2018, 40(1):1-7.
[16]
YONG P, LIAO W Y, HUANG J P, et al.Misfit function for full waveform inversion based on the Wasserstein metric with dynamic formulation[J].Journal of Computational Physics, 2019, 399:108911.
[17]
杨瑞冬, 黄建平, 杨振杰, 等.基于双对角通量校正的多尺度全波形反演[J].石油地球物理勘探, 2022, 57(5):1120-1128.YANG Ruidong, HUANG Jianping, YANG Zhenjie, et al.Multi-scale full waveform inversion based on double diagonal flux correction transport[J].Oil Geophysical Prospecting, 2022, 57(5):1120-1128.
[18]
ARAYA-POLO M, JENNINGS J, ADLER A, et al.Deep learning tomography[J].The Leading Edge, 2018, 37(1):58-66.
[19]
韩明亮,邹志辉,马锐. 利用反射地震资料和多尺度训练集的深度学习速度建模[J].石油地球物理勘探, 2021, 56(5):935-946.HAN Mingliang, ZOU Zhihui, MA Rui.Deep learning-driven velocity modeling based on seismic reflection data and multi-scale training sets[J].Oil Geophysical Prospecting, 2021, 56(5):935-946.
[20]
张兵.基于卷积神经网络和叠加速度谱的地震层速度自动建模方法[J].石油物探, 2021,60(3):366-375.ZHANG Bing.Automatic seismic interval velocity building based on convolutional neural network and velocity spectrum[J].Geophysical Prospecting for Petroleum, 2021, 60(3):366-375.
[21]
MOSSER L, KIMMAN W, DRAMSCH J, et al.Rapid seismic domain transfer:Seismic velocity inversion and modeling using deep generative neural networks[C].Extended Abstracts of 80th EAGE Conference & Exhibition, 2018, 1-5.
[22]
ZHANG Z P, WU Y, ZHOU Z, et al.VelocityGAN:subsurface velocity image estimation using conditional adversarial networks[C]. 2019 IEEE Winter Conference on Applications of Computer Vision(WACV), Waikoloa, 2019, 705-714.
[23]
FENG S H, LIN Y Z, WOHLBERG B.Multiscale data-driven seismic full-waveform inversion with field data study[J].IEEE Transactions on Geoscience and Remote Sensing, 2022, 60:1-14.
[24]
LI S C, LIU B, REN Y X, et al.Deep learning inversion of seismic data[J].IEEE Transactions on Geoscience and Remote Sensing,2020,58(3):2135-2149.
[25]
RONNEBERGER O, FISCHER P, BROX T.U-Net:Convolutional networks for biomedical image segmentation[C].Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015, Cham, 2015, 234-241.
[26]
李阳, 韩立国, 周帅, 等.基于深度学习U-net网络的重力数据界面反演方法[J]. 地球物理学报, 2023, 66(1):401-411.LI Yang, HAN Liguo, ZHOU Shuai, et al.Gravity data density interface inversion based on U-Net deep learning network[J].Chinese Journal of Geophysics, 2023, 66(1):401-411.
[27]
OKTAY O, SCHLEMPER J, FOLGOC L L, et al.Attention U-Net:Learning where to look for the pancreas [EB/OL].(2018-04-11)[2023-06-05].https://arxiv.org/abs/1804.03999.
[28]
SANDLER M, HOWARD A, ZHU M, et al.MobileNetV2:Inverted residuals and linear bottlenecks[C].2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),IEEE,Salt Lake City, 2018.
[29]
HE K M, ZHANG X Y, REN S Q, et al.Identity mappings in deep residual networks[C].Computer Vision-ECCV 2016, Cham, 2016, 630-645.
[30]
张岩, 周一帆, 宋利伟, 等.基于物理约束U-Net网络的地震数据低频延拓[J].石油地球物理勘探, 2023, 58(1):31-45.ZHANG Yan, ZHOU Yifan, SONG Liwei, et al.Low frequency continuation of seismic data based on physically constrained U-Net network[J].Oil Geophysical Prospecting, 2023, 58(1):31-45.