赵飞飞,邵亚璐,刘晓佳,闫昕蕾,韩焱
ZHAO Fei-fei,SHAO Ya-lu,LIU Xiao-jia,YAN Xin-lei,HAN Yan

• 1:中北大学信息探测与处理山西省重点实验室
• 2:中北大学省部共建动态测试技术国家重点实验室

摘要(Abstract)：

针对地下浅层地层结构复杂、地面获取的爆炸波振动信号波形混叠、频散严重，采用逆时偏移方法震源成像模糊、震源定位精度低的问题，将图像融合引入震源定位中，提出了一种基于3D-UNet的多谱图像融合定位方法。首先，对传感器采集到的信号通过变分模态分解(VMD)进行多频主成分分解，通过逆时聚焦成像方法形成多谱能量场；之后，将多谱能量场作为3D-UNet的输入，并结合注意力机制进行多谱能量场对应系数的自适应调整，通过梯度下降法使网络输出最大值位置逼近真实震源位置，生成融合网络模型。实验验证表明，本文方法相比于基于VMD多谱图像融合定位方法定位精度更高，且均方根误差在0.5 m以内。
In view of the complex structure of the shallow underground stratum, the aliasing and serious dispersion of the blast wave vibration signal obtained from the ground, using inverse time migration method to cause the problems of fuzzy seismic source imaging and low seismic location accuracy, this paper introduces image fusion into the source location, and proposes 3D-UNet-based multispectral image fusion positioning method. First, the signal collected by the sensor is decomposed into multi-frequency principal component through Variational Modal Decomposition(VMD), and then the multi-spectral energy field is formed through the reverse time focusing imaging method; After that, the multi-spectral energy field as the input of 3D-UNet, combined with the attention mechanism to adaptively adjust the coefficients of the multi-spectral energy field, so that the maximum position of the network output is continuously approaching the true focal position through the gradient descent method, and finally get the converged network model. Experimental verification shows that the method in this paper can achieve higher positioning accuracy than the multispectral image fusion positioning method based on VMD, and the root mean square error is within 0.5 m.

关键词(KeyWords)： 图像融合;3D-UNet;震源定位;VMD
image fusion;3D-UNet;vibration source location;VMD

Abstract:

Keywords:

基金项目(Foundation): 国家自然基金青年科学基金资助项目(61901419)

作者(Author): 赵飞飞,邵亚璐,刘晓佳,闫昕蕾,韩焱
ZHAO Fei-fei,SHAO Ya-lu,LIU Xiao-jia,YAN Xin-lei,HAN Yan

DOI: 10.19931/j.EB.20210397

参考文献(References)：

• [1] GEIGER L.Probability method for the determination of earthquake epicenters from arrival time only [J] .Bull St Louis Univ,1912,8(1):60-71.
• [2] WALDHAUSER F,ELLSWORTH W.A double-difference earthquake location algorithm:method and application to the northern hayward fault [J].Bulletin of the Seismological Society of America,2000,90(6):1 353-1 368.
• [3] 王泉栋，李国和，吴卫江，等.多种群遗传算法在微震振源定位中的应用[J].计算机测量与控制，2015,23(4):1 285-1 288.WANG Q D,LI G H,WU W J,et al.Application of multiple-population genetic algorithm in micor-seismic source location[J].Computer Measurement & Control,2015,23(4):1 285-1 288.
• [4] LARMAT C S,GUYER R A,JOHNSON P A.Tremor source location using time reversal:Selecting the appropriate imaging field[J].Geophysical Research Letters,2009,36(36):355-355.
• [5] ZOU Z,ZHOU H,GURROLA H.Reverse-time imaging of a doublet of microearthquakes in the three gorges reservoir region[J].Geophysical Journal International,2013,196(3):1 858-1 868.
• [6] 葛奇鑫.基于逆时成像的被动源定位与识别方法研究[D].长春：吉林大学，2019.GE Q X.Research on Passive Source Location and Identification Based on Reverse Time Imaging[D].Changchun:Jilin University,2019.
• [7] MAREN B,FRIEDEMANN W,MUSTAFA E.Preseis:A neural network-based approach to earthquake early warning for finite faults[J].Bulletin of the Seismological Society of America,2008,98(1):366-382.
• [8] 于淼.基于BP-GA混合算法的微震反演研究[D].长春：吉林大学，2013.YU M.Research on micro-seismic inversion based on BP-GA mixture algorithm[D].Changchun:Jilin University,2013.
• [9] PEROL T ,GHARBI M ,DENOLLE M.Convolutional neural network for earthquake detection and location[J].Science Advances,2018,4(2):e1700578.
• [10] 孙亚松.基于深度学习的余震P和S振相拾取研究[D].成都：成都理工大学，2019.SUN Y S.A seismic phase detection and picking method based on deep Learning[D].Chengdu:Chengdu University of Technology,2019.
• [11] 陈宇雄.全卷积神经网络地震定位方法研究[D].哈尔滨：中国地震局工程力学研究所，2019.CHEN Y X.Research of earthquake location with a fully convolutional neural network[D].Harbin:Institute of Engineering Mechanics,China Earthquake Administration,2019.
• [12] 奚先，黄江清.基于卷积神经网络的地震偏移剖面中散射体的定位和成像[J].地球物理学报，2020,63(2):687-714.XI X,HUANG J Q.Location and imaging of scatterers in seismic migration profiles based on convolution neural network[J].Chinese Journal of Geophysics,2020,63(2):687-741.
• [13] ROSS Z E,MEIER M A,HAUKSSON E.P wave arrival picking and first-motion polarity determination with deep learning[J].Journal of Geophysical Research:Solid Earth,2018,123:5 120-5 129.
• [14] 张浩，冯兴强，付昌，等.基于卷积神经网络的倾角域弹性波逆时偏移噪声压制方法[J].石油物探，2021,60(3):376-384.ZHANG H,FENG X Q,FU C,et al.Noise suppression during elastic reverse time migration in the dip-angle domain using a convolutional neural network[J].Geophysical Prospecting for Petroleum,2021,60(3):376-384.
• [15] 辛伟瑶，李剑，王小亮，等.基于深度学习的地下浅层振源定位方法[J].计算机工程，2020,46(9):292-297.XIN W Y,LI J,WANG X L,et al.Underground shallow source location method based on deep learning[J].Computer Engineering,2020,46(9):292-297.
• [16] 王小亮.基于深度学习的地下浅层振源扫描定位方法研究[D].太原：中北大学，2021.WANG X L.Research on the method of shallow underground source scanning location based on deep learning[D].Taiyuan:North University of China,2021.
• [17] 李宏，李定文，朱海琦，等.一种优化的VMD算法及其在语音信号去噪中的应用[J].吉林大学学报(理学版),2021,59(5):1 219-1 227.LI H,LI D W,ZHU H Q,et al.An optimized VMD algorithm and its application in speech signal denoising[J].Journal of Jilin University(Science edition),2021,59(5):1 219-1 277.
• [18] RONNEBERGER O,FISCHER P ,BROX T .U-Net:convolutional networks for biomedical image segmentation[J].Medical Image Computing and Computer-Assisted Intervention (MICCAI),Springer,LNCS,2015,9351:234-241.
• [19] IEK Z,ABDULKADIR A ,LIENKAMP S S ,et al.3D-UNet:Learning dense volumetric segmentation from sparse annotation[J].Springer,Cham,2016.
• [20] YAN W ,JZ B ,HC B ,et al.View adaptive learning for pancreas segmentation[J].Biomedical Signal Processing and Control,2021,66(9):102 347.
• [21] JIE H ,LI S,GANG S,et al.Squeeze-and-excitation networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2020,42(8):2 011-2 023.
• [22] CHAUDHARI S ,POLATKAN G ,RAMANATH R,et al.An attentive survey of attention models[J].ACM Transactions on Intelligent Systems and Technology 2021,1(1),Article 1.
• [23] KELVIN X,JIMMY B,RYAN K,et al.Show,attend and tell:neural image caption generation with visual attention[C]//Proceedings of the 32nd International Conference on Machine Learning,International Machine Learning Society (IMLS),Lille GrandPalais.2015,37:2 048-2 057.
• [24] WOO S,PARK J,LEE J Y,et al.CBAM:Convolutional block attention module[C]//European Conference on Computer Vision.Springer,Cham,2018.11 211.