First break picking method based on artificial intelligence and apparent velocity constraint
David COVA1,2, LIU Yang1,2,3, DING Chengzhen4, WEI Chenglin4, HU Fei4, and LI Yunzhu4
1. State Key Laboratory of Petroleum Resources and Prospecting, China University of Petroleum(Beijing), Beijing 102249, China; 2. China University of Petroleum(Beijing), Karamay Campus, Karamay, Xinjiang 834000, China; 3. CNPC Key Laboratory of Geophysical Prospecting, China University of Petroleum(Beijing), Beijing 102249, China; 4. Geophysical Research Institute, BGP Inc., CNPC, Zhuozhou, Hebei 072751, China
Abstract:Picking seismic first breaks is an important step for correcting near-surface long-wavelength static anomalies.Nowadays,dense acquisition brings exponentially increasing seismic data, so that it is necessary to find a new method to pick first breaks.Conventional methods rely on manual picking and quality control, which is inefficient for large datasets.Compared with conventional methods, deep learning can greatly improve picking efficiency.Among the deep learning algorithms for picking first breaks, Fully Convolutional Networks (FCNs) have outstanding advantages in semantic segmentation, they can process data with variable sizes and perform high-resolution pixel classification.However, such segmentation has shortcomings in locating accuracy.U-Net is a variant of FCN that can solve the problem of first break picking.Although it is characterized by easy implementation,the accuracy decreases when the signal-to-noise ratio is low.In order to eliminate the limitation, this paper proposes four key points:(1) Design a workflow to balance the amplitude of samples,thus improving the network accuracy; (2) Compare three state-of-the-art U-Net variants with varying complexity, including Wide U-Net, UNet++, and Attention U-Net; (3) Optimize the network's hyperparameters with categorical loss and improved activation functions;(4) Use apparent velocity to constrain and improve the segmentation accuracy. Comparison of U-Net and its variants with different complexity has shown that U-Net has the best accuracy and efficiency. Finally, the results over a land dataset are promising.
作者简介: David Cova,博士研究生,1986年生;2008年毕业于委内瑞拉西蒙·玻利瓦尔大学,获工程地球物理专业学士学位,2018年获中国石油大学(北京)地球物理专业硕士学位;现为中国石油大学(北京)在读博士生,主要研究方向为地震属性、油气储层智能预测等。
引用本文:
David Cova, 刘洋, 丁成震, 魏程霖, 胡飞, 李韵竹. 人工智能和视速度约束的地震波初至拾取方法[J]. 石油地球物理勘探, 2021, 56(3): 419-435.
David COVA, LIU Yang, DING Chengzhen, WEI Chenglin, HU Fei, and LI Yunzhu. First break picking method based on artificial intelligence and apparent velocity constraint. Oil Geophysical Prospecting, 2021, 56(3): 419-435.
刘百红,李建华,郑四连,等. 基于局部相似属性的初至自动拾取[J].石油地球物理勘探,2018,53(3):449-453.LIU Baihong,LI Jianhua,ZHENG Silian,et al.Automatic first break picking based on local similarity attribute[J].Oil Geophysical Prospecting,2018,53(3):449-453.
[2]
宋建国,李赋真,徐维秀,等.改进的神经网络级联相关算法及其在初至拾取中的应用[J].石油地球物理勘探,2018,53(1):8-16.SONG Jianguo,LI Fuzhen,XU Weixiu,et al.An improved neural-network cascade-correlation algorithm and its application in seismic first break picking[J].Oil Geophysical Prospecting,2018,53(1):8-16.
[3]
刘腾蛟,高阳,储仿东,等.最小二乘曲线拟合的微地震初至优化拾取方法及应用[J].石油地球物理勘探, 2018,53(增刊2):124-129.LIU Tengjiao,GAO Yang,CHU Fangdong,et al. Microseismic first arrival picking based on least square curve fitting[J].Oil Geophysical Prospecting,2018,53(S2):124-129.
[4]
胡瑞卿,王彦春,尹志恒,等.结合CEEMDAN和主成分分析的低信噪比微地震初至信号检测[J].石油地球物理勘探,2019,54(1):45-53.HU Ruiqing,WANG Yanchun,YIN Zhiheng,et al.A first arrival detection method in low SNR microseismic signals based on CEEMDAN-PCA[J].Oil Geophysical Prospecting,2019,54(1):45-53.
[5]
Goodfellow I,Bengio Y,Courville A C.Deep learning[J].Nature,2015,521(7553):436-444.
[6]
Bengio Y.Deep learning of representations for unsupervised and transfer learning[C].ICML Unsupervised and Transfer Learning,2012.
[7]
庄东海,肖春燕,颜永宁.利用人工神经网络自动拾取地震记录初至[J].石油地球物理勘探,1994,29(5):659-664.ZHUANG Donghai,XIAO Chunyan,YAN Yongning. Seismic first arrivel pickup using artificial neural network[J].Oil Geophysical Prospecting,1994,29(5):659-664.
[8]
曹晓莉,刘斌,王淑荣,等. 综合动量法和可变学习速度的BP神经网络地震初至拾取[J].石油地球物理勘探,2020,55(1):71-79.CAO Xiaoli,LIU Bin,WANG Shurong,et al.Seismic first-break picking based on BP neural network integrated with momentum method and adaptive learning rate method[J].Oil Geophysical Prospecting,2020,55(1):71-79.
[9]
蒋一然,宁杰远.基于支持向量机的地震体波震相自动识别及到时自动拾取[J].地球物理学报,2019,62(1):361-373.JIANG Yiran,NING Jieyuan.Automatic detection of seismic body-wave phases and determination of their arrival times based on support vector machine[J].Chinese Journal of Geophysics,2019,62(1):361-373.
[10]
谭家炜,李静,李飞达,等.基于超虚干涉法约束的模糊C均值聚类地震初至自动提取[J].石油地球物理勘探,2020,55(5):979-990.TAN Jiawei,LI Jing,LI Feida,et al.Automatic pick-up of seismic P-wave first arrivals via fuzzy C-means method constrained by super-virtual interferometry[J].Oil Geophysical Prospecting,2020,55(5):979-990.
[11]
王维波,徐西龙,盛立,等.卷积神经网络微地震事件检测[J].石油地球物理勘探,2020,55(5):939-949.WANG Weibo,XU Xilong,SHENG Li,et al.Detection of microseismic events based on convolutional neural network[J].Oil Geophysical Prospecting,2020,55(5):939-949.
[12]
Hollander Y,Merouane A,Yilmaz O.Using a deep convolutional neural network to enhance the accuracy of first-break picking[C].SEG Technical Program Expanded Abstracts,2018,37:4628-4632.
[13]
Yuan S,Liu J,Wang S,et al.Seismic waveform classification and first break picking using convolution neural networks[J].IEEE Geoscience & Remote Sensing Letters,2018,15(2):272-276.
[14]
Duan X,Zhang J,Liu Z,et al.Integrating seismic first-break picking methods with a machine learning approach[C].SEG Technical Program Expanded Abstracts,2018,37:2186-2190.
[15]
Tsai K C,Hu W,Wu X,et al.First-break automatic picking with deep semisupervised learning neural network[C].SEG Technical Program Expanded Abstracts,2018,37:2181-2185.
[16]
李薇薇,龚仁彬,周相广,等.基于深度学习UNet++网络的初至波拾取方法[J].地球物理学进展,2021,36(1):187-194.LI Weiwei,GONG Renbin,ZHOU Xiangguang,et al.UNet++:a deep-neural-network-based seismic arrival time picking method[J].Progress in Geophysics,2021,36(1):187-194.
[17]
赵明,陈石,房立华,等.基于U形卷积神经网络的震相识别与到时拾取方法研究[J].地球物理学报,2019,62(8):3034-3042.ZHAO Ming,CHEN Shi,FANG Lihua,et al.Earthquake phase arrival auto-picking based on U-shaped convolutional neural network[J].Chinese Journal of Geophysics,2019,62(8):3034-3042.
[18]
陈德武,杨午阳,魏新建,等.基于混合网络U-SegNet的地震初至自动拾取[J].石油地球物理勘探,2020,55(6):1188-1201.CHEN Dewu,YANG Wuyang,WEI Xinjian,et al.Automatic picking of seismic first arrivals based on hybrid network U-SegNet[J].Oil Geophysical Prospecting,2020,55(6):1188-1201.
[19]
Xie T,Yue Z,Jiao X,et al.First-break automatic pi-cking with fully convolutional networks and transfer learning[C].SEG Technical Program Expanded Abstracts,2019,38:4972-4976.
[20]
LeCun Y,Boser B,Denker J S,et al.Backpropagation applied to handwritten zip code recognition[J].Neural Computation,1989,1(4):541-551.
[21]
Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks:Advance in neural information processing systems[C].NIPS,2012,1097-1105.
[22]
Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[C].International Conference on Learning Representations,2015.
[23]
Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C].IEEE Conference on Computer Vision and Pattern Recognition,2015:1-9.
[24]
Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,39(4):640-651.
[25]
Noh H,Hong S,Han B.Learning deconvolution network for semantic segmentation[C].IEEE International Conference on Computer Vision,2015,1520-1528.
[26]
Chen L C,Papandreou G,Kokkinos I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2018,40(4):834-848.
[27]
Yu F,Koltun V.Multi-scale context aggregation by dila-ted convolutions[C].International Conference on Lear-ning Representations,2015.
[28]
Ronneberger O,Fischer P,Brox T.U-Net:Convolutional networks for biomedical image segmentation[C].International Conference on Medical Image Computing and Computer-Assisted Intervention,2015.
[29]
Hu L,Zheng X,Duan Y.U-Net convolutional networks for first arrival picking[C].SEG Technical Program Expanded Abstracts,2018,37:15-18.
[30]
Zhou Z,Siddiquee M M,Tajbakhsh N,et al.UNet++:Redesigning skip connections to exploit multisdcal features in image segmentation[J].IEEE Transactions on Medical Imaging, 2020,39:1856-1867.
[31]
Abraham N,Khan N M.a novel Focal Tversky loss function with improved attention U-Net for lesion segmentation[C].IEEE 16th International Sympo-sium on Biomedical Imaging,2019,683-687.
[32]
Ma Y,Cao S,Rector J W,et al.Automatic first arrival picking for borehole seismic data using a pixel-level network[C].SEG Technical Program Expanded Abstracts,2019,38:2463-2467.
Sudre C H,Li W,Vercauteren T,et al.Generalised Dice overlap as a deep learning loss function for highly unba-lanced segmentations[C].Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support,2017,240-248.
[36]
Tversky A.Features of similarity[J].Psychological Review,1977,84(4):327-352.
[37]
Hashemi S R,Seyed S M S,Erdogmus D,et al.Asymmetric loss functions and deep densely-connected networks for highly-imbalanced medical image segmentation:Application to multiple sclerosis lesion detection[J].IEEE Access,2019,7(1):1721-1735.
[38]
Huber P J.Robust regression:Asymptotics,conjectures and Monte Carlo[J].Annals of Statistics,1973,1(5):799-821.
[39]
Upton G,Cook I.Understanding Statistics[M].Oxford University Press,1996.
[40]
Basheer I A,Hajmeer M N.Artificial neural networks:fundamentals,computing,design,and application[J].Journal of Microbiological Methods,2000,43(1):3-31.
[41]
Berman M, Triki A R, Blaschko M B.The Lovasz-softmax loss:A tractable surrogate for the optimization of the intersection-over-union measure in neural networks[C].IEEE/CVF Conference on Computer Vision and Pattern Recognition,2018,4413-4421.
[42]
Lu K,Feng S.Auto-windowed super-virtual interfe-rometry via machine learning:A strategy of first-arrival traveltime automatic picking for noisy seismic data[C].SEG Technical Program Expanded Abstracts,2018,37:10-14.
[43]
Dosovitskiy A,Fischer P,Springenberg J T,et al.Discriminative unsupervised feature learning with exemplar convolutional neural networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(9):1734-1747.
[44]
Robbins H,Monro S A.Stochastic approximation method[J].The Annals of Mathematical Statistics,1951,22(3):400-407.
[45]
Zeiler M D.ADADELTA:An adaptive learning rate method[J].ArXiv Preprint ArXiv:1212.5701,2012.
[46]
Kingma D,Ba J.Adam:A method for stochastic optimization[C].3rd International Conference on Learning Representations,2015.
[47]
LeCun Y,Bottou L,Orr G,et al.Efficient BackProp//Neural Networks:Tricks of the Trade[M].Springer,1998.
[48]
Glorot X,Bengio Y.Understanding the difficulty of trai-ning deep feedforward neural networks[C].Thirteenth International Conference on Artificial Intelligence and Statistics,JMLR Workshop and Conference Proceedings,2010,249-256.
[49]
He K, Zhang X, Ren S,et al.Delving deep into rectifiers:surpassing human-level performance on image Net classification[C].IEEE International Conference on Computer Vision(ICCV),2015,1026-1034.
[50]
Zou K H,O'Malley A J,Mauri L.Receiver-operating characteristic analysis for evaluating diagnostic tests and predictive models[J].Circulation,2007,115(5):654-657.