Please wait a minute...
浙江大学学报(工学版)  2020, Vol. 54 Issue (5): 978-984    DOI: 10.3785/j.issn.1008-973X.2020.05.016
地球科学     
基于卷积自编码器的地震数据处理
江金生(),任浩然*(),李瀚野
浙江大学 地球科学学院,浙江省地学大数据与地球深部资源重点实验室,浙江 杭州 310027
Seismic data processing based on convolutional autoencoder
Jin-sheng JIANG(),Hao-ran REN*(),Han-ye LI
Key Laboratory of Geoscience Big Data and Deep Resource of Zhejiang Province, School of Earth Sciences, Zhejiang University, Hangzhou 310027, China
 全文: PDF(1052 KB)   HTML
摘要:

引入以深度学习为代表的数据驱动方法,加速地震数据的处理流程,获得更精确的地下介质信息. 卷积自编码器方法在地震数据压缩降维的同时,利用数据的空间局部相关性自动提取信号特征,避免数学物理模型的假设依赖. 通过设计不同地质模型的地下速度结构,利用波动方程正演模拟构建大量不同特征的地震数据训练集和测试集. 与模型驱动的地震随机噪声压制和地震道插值方法不同,数据驱动下的卷积自编码器方法能够从含随机噪声地震数据和地震道缺失数据中,直接识别和提取出其中的有效地震信号,从而压制随机噪声以及重建原始地震数据,实验结果验证了该方法的有效性. 卷积自编码器方法不需要人工阈值控制,具有更高的处理效率.

关键词: 卷积自编码器地震道插值地震随机噪声压制深度神经网络稀疏表达    
Abstract:

Data-driven methods represented by deep learning were introduced into seismic data processing, which can speed up the processing process of seismic data and obtain more accurate information about underground media. The convolutional autoencoder (CAE) method can extract data features by using the spatial correlation of seismic data while compressing it, which avoids the assumption dependence of mathematical and physical model. The underground velocity structures of different geological models were designed, and the training and test sets with different features were obtained through wave equation forward modeling. CAE method is different from the traditional model-driven method of seismic random noise attenuation and seismic data interpolation. CAE method can identify and extract the valid seismic signals directly from seismic data with random noise and missing traces, to attenuate random noise and reconstruct the original seismic data. Experimental results verified the effectiveness of this method. CAE method gets rid of the artificial threshold, therefore has higher processing efficiency.

Key words: convolutional autoencoder    seismic data interpolation    seismic random noise attenuation    deep neural network    sparse representation
收稿日期: 2019-06-11 出版日期: 2020-05-05
CLC:  P 3  
基金资助: 国家自然科学基金资助项目(41674123);浙江省自然科学基金资助项目(LY19D040002);中石化地球物理重点实验室开放基金资助项目;多方法大数据智能反演技术及软件研发基金资助项目(2018YFC0603604)
通讯作者: 任浩然     E-mail: jiangjs217@zju.edu.cn;rhr@zju.edu.cn
作者简介: 江金生(1994—),男,硕士生,从事地球物理信号处理与人工智能研究. orcid.org/0000-0002-3653-2055. E-mail: jiangjs217@zju.edu.cn
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章  
江金生
任浩然
李瀚野

引用本文:

江金生,任浩然,李瀚野. 基于卷积自编码器的地震数据处理[J]. 浙江大学学报(工学版), 2020, 54(5): 978-984.

Jin-sheng JIANG,Hao-ran REN,Han-ye LI. Seismic data processing based on convolutional autoencoder. Journal of ZheJiang University (Engineering Science), 2020, 54(5): 978-984.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2020.05.016        http://www.zjujournals.com/eng/CN/Y2020/V54/I5/978

图 1  卷积自编码器网络架构
图 2  卷积自编码器的训练和测试流程
图 3  用于产生训练集的地下速度模型
图 4  地震随机噪声压制结果比对
图 5  地震数据插值重建结果比对
1 张英强, 陆文凯.基于迭代收缩高分辨率Radon变换的地震道插值[C]//中国地球物理学会第二十九届年会.昆明: 中国地球物理学会, 2013: 632-633.
ZHANG Ying-qiang, LU Wen-kai. Seismic trace interpolation method based on iterative shrinking high-resolution radon transform [C]// The 29th Annual Meeting of the Chinese Geophysical Society. Kunming: Chinese Geophysical Society, 2013: 632-633.
2 谭军, 宋鹏, 李金山, 等 基于速度加权叠加和AVO分析的叠前地震数据插值方法[J]. 石油物探, 2017, 56 (6): 812- 819
TAN Jun, SONG Peng, LI Jin-shan, et al Pre-stack seismic data interpolation based on velocity-weighted stack and AVO analysis[J]. Geophysical Prospecting for Petroleum, 2017, 56 (6): 812- 819
doi: 10.3969/j.issn.1000-1441.2017.06.006
3 王立明, 刘保华, 许江, 等 一种有效的地震道插值方法[J]. 地球物理学进展, 2012, 27 (6): 2561- 2569
WANG Li-ming, LIU Bao-hua, XU Jiang, et al An effective seismic trace interpolation method[J]. Progressin Geophys, 2012, 27 (6): 2561- 2569
doi: 10.6038/j.issn.1004-2903.2012.06.033
4 SHAN H, MA J, YANG H Comparisons of wavelets, contourlets and curvelets in seismic denoising[J]. Journal of Applied Geophysics, 2009, 69 (2): 103- 115
doi: 10.1016/j.jappgeo.2009.08.002
5 薛昭, 董良国, 单联瑜 Radon变换去噪方法的保幅性理论分析[J]. 石油地球物理勘探, 2012, 47 (6): 858- 867
XUE Zhao, DONG Liang-guo, SHAN Lian-yu Amplitude preservation theoretical analysis of radon transforms denoising method[J]. OilGeophysical Prospecting, 2012, 47 (6): 858- 867
6 彭才, 常智, 朱仕军 基于曲波变换的地震数据去噪方法[J]. 石油物探, 2008, 47 (5): 461- 464
PENG Cai, CHANG Zhi, ZHU Shi-jun Noise elimination method based on curvelet transform[J]. Geophysical Prospecting for Petroleum, 2008, 47 (5): 461- 464
doi: 10.3969/j.issn.1000-1441.2008.05.006
7 吴招才, 刘天佑 地震数据去噪中的小波方法[J]. 地球物理学进展, 2008, 23 (2): 493- 499
WU Zhao-cai, LIU Tian-you Wavelet transform methods in seismic data noise attenuation[J]. Progress in Geophysics, 2008, 23 (2): 493- 499
8 颜中辉, 栾锡武, 王赟, 等 基于经验模态分解的分数维地震随机噪声衰减方法[J]. 地球物理学报, 2017, 60 (7): 2845- 2857
YAN Zhong-hui, LUAN Xi-wu, WANG Yun, et al Seismic random noise attenuation based on empirical mode decomposition of fractal dimension[J]. Chinese Journal of Geophysics, 2017, 60 (7): 2845- 2857
doi: 10.6038/cjg20170729
9 NEELAMANI R, BAUMSTEIN A I, GILLARD D G, et al Coherent and random noise attenuation using the curvelet transform[J]. The Leading Edge, 2008, 27: 240- 248
doi: 10.1190/1.2840373
10 HINTON G, DENG L, YU D, et al Deep neural networks for acoustic modeling in speech recognition[J]. IEEE Signal Processing Magazine, 2012, 29 (6): 82- 97
doi: 10.1109/MSP.2012.2205597
11 LECUN Y, BENGIO Y, HINTON G Deep learning[J]. Nature, 2015, 521 (7553): 436
doi: 10.1038/nature14539
12 GUO Y, LIU Y, OERLEMANS A, et al Deep learning for visual understanding: a review[J]. Neurocomputing, 2016, 187: 27- 48
doi: 10.1016/j.neucom.2015.09.116
13 GRAMSTAD O, NICKELM. Automated interpretation of top and base salt using deep-convolutional networks [C]// 88th Annual International Meeting, SEG. Anaheim: Society of Exploration Geophysicists, 2018: 1956-1960.
14 WU X, SHI Y, FOMEL S. Convolutional neural networks for fault interpretation in seismic images [C]// 88th Annual International Meeting, SEG. Anaheim: Society of Exploration Geophysicists, 2018: 1946-1950.
15 ALFARRAJ M, ALREGIB G. Petrophysicalproperty estimation from seismic data using recurrent neural networks [C]// 88th Annual International Meeting, SEG. Anaheim: Society of Exploration Geophysicists, 2018: 2141-2146.
16 ALWON S. Generative adversarial networks in seismic data processing [C]// 88th Annual International Meeting, SEG. Anaheim: Society of Exploration Geophysicists, 2018: 1991-1995.
17 LI H, YANG W, YONG X. Deep learning for ground-roll noise attenuation [C]// 88th Annual International Meeting, SEG. Anaheim: Society of Exploration Geophysicists, 2018: 1981-1985.
18 王钰清, 陆文凯, 刘金林, 等 基于数据增广和CNN的地震随机噪声压制[J]. 地球物理学报, 2019, 62 (1): 421- 433
WANG Yu-qing, LU Wen-kai, LIU Jin-lin, et al Random seismic noise attenuation based on data augmentation and CNN[J]. Chinese Journal of Geophysics, 2019, 62 (1): 421- 433
doi: 10.6038/cjg2019M0385
19 HINTON G E, SALAKHUTDINOV R R Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313 (5786): 504- 507
doi: 10.1126/science.1127647
20 SHELHAMER E, LONG J, DARRELL T Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (4): 640- 651
doi: 10.1109/TPAMI.2016.2572683
21 BADRINARAYANAN V, KENDALL A, CIPOLLA R SegNet: adeep convolutional encoder-decoder aqrchitecture for scene segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39 (12): 2481- 2495
doi: 10.1109/TPAMI.2016.2644615
22 贾文娟, 张煜东 自编码器理论与方法综述[J]. 计算机系统应用, 2018, 27 (5): 1- 9
JIA Wen-juan, ZHANG Yu-dong Survey on theories and methods of autoencoder[J]. Computer Science and Applications, 2018, 27 (5): 1- 9
23 MASCI J, MEIER U, CIRESAN D, et al. Stacked convolutional auto-encoders for hierarchical feature extraction [C]// Artificial Neural Networks and Machine Learning–ICANN 2011. Berlin, Heidelberg: Springer Berlin Heidelberg, 2011: 52-59.
24 BENGIO Y, COURVILLE A, VINCENT P Representation learning: areview and new perspectives[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35 (8): 1798- 1828
doi: 10.1109/TPAMI.2013.50
25 刘洋, 李炳秀, 王典, 等 基于正则化条件的地震数据局部信噪比估计方法[J]. 地球物理学报, 2017, 60 (5): 1979- 1987
LIU Yang, LI Bing-xiu, WANG Dian, et al Local SNR estimation method based on regularization for seismic data[J]. Chinese Journal of Geophysics, 2017, 60 (5): 1979- 1987
doi: 10.6038/cjg20170529
[1] 许佳辉,王敬昌,陈岭,吴勇. 基于图神经网络的地表水水质预测模型[J]. 浙江大学学报(工学版), 2021, 55(4): 601-607.
[2] 毛奕喆,龚国芳,周星海,王飞. 基于马尔可夫过程和深度神经网络的TBM围岩识别[J]. 浙江大学学报(工学版), 2021, 55(3): 448-454.
[3] 陈世达,刘强,韩亮. 降低分布式训练通信的梯度稀疏压缩方法[J]. 浙江大学学报(工学版), 2021, 55(2): 386-394.
[4] 陈纬奇,王敬昌,陈岭,杨勇勤,吴勇. 基于深度神经网络的多因素感知终端换机预测模型[J]. 浙江大学学报(工学版), 2021, 55(1): 109-115.
[5] 王昶,张永生,王旭,于英. 基于深度学习的遥感影像变化检测方法[J]. 浙江大学学报(工学版), 2020, 54(11): 2138-2148.
[6] 孙伟伟, 马俊, 杨刚, 李巍岳. 改进核空间对称稀疏表达用于高光谱波段选择[J]. 浙江大学学报(工学版), 2018, 52(4): 687-693.