基于电磁卫星的闪电哨声波智能检测算法的研究进展

袁静, 王桥, 张学民, 杨德贺, 王志国, 张乐, 申旭辉, 泽仁志玛. 2021. 基于电磁卫星的闪电哨声波智能检测算法的研究进展. 地球物理学报, 64(5): 1471-1495, doi: 10.6038/cjg2021O0263
引用本文: 袁静, 王桥, 张学民, 杨德贺, 王志国, 张乐, 申旭辉, 泽仁志玛. 2021. 基于电磁卫星的闪电哨声波智能检测算法的研究进展. 地球物理学报, 64(5): 1471-1495, doi: 10.6038/cjg2021O0263
YUAN Jing, WANG Qiao, ZHANG XueMin, YANG DeHe, WANG ZhiGuo, ZHANG Le, SHEN XuHui, Zeren Zima. 2021. Advances in the automatic detection algorithms for lightning whistlers recorded by electromagnetic satellite data. Chinese Journal of Geophysics (in Chinese), 64(5): 1471-1495, doi: 10.6038/cjg2021O0263
Citation: YUAN Jing, WANG Qiao, ZHANG XueMin, YANG DeHe, WANG ZhiGuo, ZHANG Le, SHEN XuHui, Zeren Zima. 2021. Advances in the automatic detection algorithms for lightning whistlers recorded by electromagnetic satellite data. Chinese Journal of Geophysics (in Chinese), 64(5): 1471-1495, doi: 10.6038/cjg2021O0263

基于电磁卫星的闪电哨声波智能检测算法的研究进展

  • 基金项目:

    中央直属高校基本科研业务经费(ZY20180122), 中国科技部国家重点研发计划(2018YFC1503502和2018YFC1503806和2018YFC1503501), 廊坊科技局科学研究与发展计划自筹经费项目(2020011025), 中央直属高校基本科研业务经费(2020011025), APSCO earthquake project II: integrating satellite and ground-based observation for earthquake precursors and signatures(亚太地震二期项目: 地震前兆特征的星地一体化观测研究)项目资助, ISSI-BJ(2019IT-33)项目资助

详细信息
    作者简介:

    袁静, 女, 1981年出生, 副教授, 博士, 2019年毕业于清华大学, 主要从事大数据挖掘技术方面的研究.E-mail: yuanjing20110824@sina.com

    通讯作者: 杨德贺, 男, 1985年出生, 助理研究员, 博士, 2014年毕业于中国矿业大学(北京), 主要从事大数据挖掘技术方面的研究.E-mail: ydhmmm@163.com
  • 中图分类号: P352

Advances in the automatic detection algorithms for lightning whistlers recorded by electromagnetic satellite data

More Information
  • 闪电哨声波作为探索空间物理环境的重要媒介,淹没在海量的电磁卫星数据中.近年来随着计算机视觉和深度学习等人工智能技术的发展,从电磁卫星的存档数据中自动检测闪电哨声波的算法取得了一定的效果.本文对近年来闪电哨声波智能检测算法的文献进行了整理和总结.首先,阐述闪电哨声波在电磁卫星数据中呈现的时频特征和类型;然后,介绍了闪电哨声波智能检测算法的流程并从闪电哨声波的特征提取、分类和定位三个方面对主要的智能检测算法进行归纳、总结和评述;其次,简述了闪电哨声波智能检测模型的评价指标;接着,在张衡一号(ZH-1)卫星的磁场数据上对三种典型的闪电哨声波智能检测算法进行复现,并对三种算法的优缺点进行了较深入的分析;最后,对基于电磁卫星的闪电哨声波智能检测的研究领域进行总结和展望.

  • 加载中
  • 图 1 

    闪电哨声波传播路径

    Figure 1. 

    Propagation path of a lightning whistler

    图 2 

    张衡一号卫星闪电哨声波形态图例

    Figure 2. 

    The time-frequency spectrogram of the lightning whistler observed by ZH-1 satellite

    图 3 

    DEMETER卫星闪电形态图例(Parrot et al., 2015)

    Figure 3. 

    The time-frequency spectrogram of the lightning whistler observed by DEMETER (Parrot et al., 2015)

    图 4  2019年7月28日全球闪电密度分布图(http://dudwlln.otago.ac.nz/)

    Figure 4.  A global lightning density map on July 28, 2019 (http://dudwlln.otago.ac.nz/)

    图 5 

    闪电哨声波智能检测的流程

    Figure 5. 

    The overall flow chart of the lightning whistlers detection

    图 6 

    张衡一号感应式磁力仪测得的快速变化磁场数据

    Figure 6. 

    The fast variation data of geomagnetic field obtained by Search Coil Magnetometer onboard ZH-1 satellite

    图 7 

    张衡一号感应式磁力仪快速变化磁场数据的时频图

    Figure 7. 

    The time-frequency spectrogram of the Search Coil Magnetometer data in Fig. 6

    图 8 

    时频图(The Stanford VLF Group, 2018)

    Figure 8. 

    The time-frequency spectrogram (The Stanford VLF Group, 2018)

    图 9 

    去噪和网格化(The Stanford VLF Group, 2018)

    Figure 9. 

    Denoising and meshing (The Stanford VLF Group, 2018)

    图 10 

    合并(The Stanford VLF Group, 2018)

    Figure 10. 

    Binning (The Stanford VLF Group, 2018)

    图 11 

    灰度化的时频图(Dharma et al., 2014)

    Figure 11. 

    Grayscale image of the time-frequency spectrogram (Dharma et al., 2014)

    图 12 

    二值化处理后的时频图(Dharma et al., 2014)

    Figure 12. 

    Binary image of the time-frequency spectrogram (Dharma et al., 2014)

    图 13 

    中值滤波和开运算处理后的时频图(Dharma et al., 2014)

    Figure 13. 

    The results by median filter and open operation (Dharma et al., 2014)

    图 14 

    连通域标记(Dharma et al., 2014)

    Figure 14. 

    Labeling the connected domain (Dharma et al., 2014)

    图 15 

    滑动的深度卷积神经网络(Konan et al., 2020)

    Figure 15. 

    Sliding deep convolutional neural networks (Konan et al., 2020)

    图 16 

    卷积层的输出(Konan et al., 2020)

    Figure 16. 

    Output of convolutional layer (Konan et al., 2020)

    图 17 

    YOLOV3的网络结构(Redmon and Farhadi, 2018)

    Figure 17. 

    The network structure of YOLOV3 (Redmon and Farhadi, 2018)

    图 18 

    YOLOV3的残差网络(Redmon and Farhadi, 2018)

    Figure 18. 

    The residual network in the network structure of YOLOV3 (Redmon and Farhadi, 2018)

    图 19 

    闪电哨声波模板(Fiser et al., 2010)

    Figure 19. 

    Lightning whistler template (Fiser et al., 2010)

    图 20 

    基于互相关熵的闪电哨声波自动检测(Fiser et al., 2010)

    Figure 20. 

    Automatic detection of lightning whistler with cross-correlation entropy (Fiser et al., 2010)

    图 21 

    时间变化与频率之间的关系图(Oike et al., 2014)

    Figure 21. 

    Relationship between time and frequency (Oike et al., 2014)

    图 22 

    检测闪电哨声波的结果(Oike et al., 2014)

    Figure 22. 

    Detection results of lightning whistler (Oike et al., 2014)

    图 23 

    决策规则(Ali Ahmad et al., 2019)

    Figure 23. 

    Decision Rules (Ali Ahmad et al., 2019)

    图 24 

    滑动的三层深度卷积神经网络(Konan et al., 2020)

    Figure 24. 

    Sliding three-layer deep convolutional neural network (Konan et al., 2020)

    图 25 

    FPN特征融合:(a) 13×13;(b) 26×26;(c) 52×52.

    Figure 25. 

    FPN fusion

    图 26 

    YOLOV3坐标模型(Redmon and Farhadi, 2018)

    Figure 26. 

    Coordinate representation in mathematical model of the YOLOV3 (Redmon and Farhadi, 2018)

    图 27 

    旋转不变性

    Figure 27. 

    Rotation invariance

    图 28 

    实例

    Figure 28. 

    An example

    图 29 

    三种代表性的算法

    Figure 29. 

    Three typical algorithms

    图 30 

    初始学习率0.1,轮次40

    Figure 30. 

    Learning Rate=0.1 and epoch=40

    图 31 

    初始学习率0.01,轮次40

    Figure 31. 

    Learning Rate=0.01 and epoch=40

    图 32 

    初始学习率0.001,轮次40

    Figure 32. 

    Learning Rate=0.001 and epoch=40

    图 33 

    初始学习率0.0001,轮次40

    Figure 33. 

    Learning Rate=0.0001 and epoch=40

    图 34 

    初始学习率0.001,轮次70

    Figure 34. 

    The learning rate=0.001 and epoch=70

    图 35 

    神经网络的训练过程

    Figure 35. 

    Training the neural networks

    图 36 

    基于形态学处理的闪电哨声波检测算法

    Figure 36. 

    The Process of lightning whistler detection algorithm based on morphological processing

    图 37 

    滑动卷积神经网络的检测的效果

    Figure 37. 

    SDNN neural network

    图 38 

    YOLOV3神经网络的检测的效果

    Figure 38. 

    YOLOV3 neural network

    图 39 

    基于哈希的哨声波图像智能检索

    Figure 39. 

    The image retrieval technology of lightning whistler based on Hashing

    表 1 

    闪电哨声波8种分类(Helliwell, 1965)

    Table 1. 

    Type and principal definitions of lightning whistlers (Helliwell, 1965)

    类别 英文名称 时频特征描述 时频特征图
    单跳 One Hop 哨声波持续时间大约1 s(如右图L色散线所示,竖线代表闪电源)
    两跳 Two Hop 哨声波持续的时间大约2s(如右图L色散线所示,竖线代表闪电源)
    混合跳 Hybrid 单跳和两跳混合(如右图L色散线所示,竖线代表闪电源)
    回波 Echo Train 一组哨声波,组内的哨声波的降频幅度越来越小,分为奇数回波和偶数回波.
    奇数回波指的组内的哨声波之间的时间间隔是1:3:5:7;偶数回波指的组内的哨声波之间的时间间隔是2:4:6:8;(如右图L色散线所示,竖线代表闪电源)
    多成分 Multiple-component 多路径:一个闪电激发电磁波经过不同的传输路径.
    混合路径:单跳的哨声波路径中含有双跳或多跳的哨声.(如右图L色散线所示,竖线代表闪电源)
    多闪电源 Multiple-source 多个闪电源激发的闪电哨声波.(如右图L色散线所示,竖线代表闪电源)
    鼻型 Nose 哨声波存在两种时频趋势:高于鼻频(圆圈位置为鼻频),哨声波的频率随着时间的推移而增加.同时鼻频下,频率随着时间的推移而降低.(如右图L色散线所示,竖线代表闪电源)
    分数跳 Fractional Hop 闪电哨声波持续的时间小于1 s.(如右图L色散线所示,竖线代表闪电源)
    下载: 导出CSV

    表 2 

    矩阵

    Table 2. 

    Matrix

    预测样本的类别
    1 0
    样本的实际类别 1 真正例(True Positive, TP)
    实际的样本类别:P
    预测的类别: P
    假负例(False Negative, FN)
    实际的样本类别:P
    预测的类别: N
    0 假正例(False Positive, FP)
    实际的样本类别:N
    预测的类别: P
    真负例(True Negative,TN)
    实际的样本类别:N
    预测的类别: N
    下载: 导出CSV

    表 3 

    闪电哨声波检测算法的性能指标比较

    Table 3. 

    Performance indexes for lightning whistler detection algorithm

    方法 错分率 虚警率 F1值 数据说明 基准数据
    斯坦福团队(2018)提供的算法 - - - 用DEMETER卫星的电场数据只用来检测持续时间0.2 s左右具有分数跳类型的闪电哨声波. 该算法发布在斯坦福大学的甚低频团队的项目组的网站上,未提供算法性能评估指标和评估结果.
    Fiser等(2010)提出的检测算法 - 48.6% - 用DEMETER卫星的电场数据, 只用来检测具有分数跳类型的闪电哨声波, 共1192个闪电哨声波事件. 欧洲闪电探测网络(http://www.euclid.org)
    Oike等(2014)提出的检测算法 - 4.4% - AkebonoX射线观测卫星的VLF载荷数据 全球闪电定位网络(http://www.wwlln.com/)
    Ali Ahmad等(2019)提出的检测算法 25.07% - - Arase卫星的观测到的磁场数据 全球闪电定位网络(http://www.wwlln.com/)
    Konan等(2020)提供的滑动卷积神经网络的检测算法 9.3% 2.6% 93.9% 收集地面马里恩岛观测站的2196个样本数据 用AWD算法检测的结果做基准数据(Lichtenberger et al., 2008).
    Konan等(2020)提供的YOLOV3目标检测神经网络 14.1% 2.1% 91.5% 收集地面马里恩岛观测站的2196个样本数据 用AWD算法检测的结果做基准数据(Lichtenberger et al., 2008).
    下载: 导出CSV

    表 4 

    SDNN网络的调参过程

    Table 4. 

    Parameter adjustment of the SDNN model

    学习率 优化器 批次大小 激活函数 轮次 最终的loss
    0.1 Adam 64 Relu 40 1.087
    0.01 Adam 64 Relu 40 1.051
    0.001 Adam 64 Relu 40 0.157
    0.0001 Adam 64 Relu 40 0.384
    0.001 Adam 64 Relu 70 0.0603
    0.001 Adam 128 Relu 70 0.0714
    0.001 Adam 256 Relu 70 0.0742
    0.001 Adam 32 Relu 70 0.0089
    0.001 Adam 32 tanh 70 1.703
    0.001 Adam 32 Softmax 70 1.847
    0.001 Adam 32 Sigmoid 70 1.912
    0.001 RMSprop 32 Relu 70 0.9742
    0.001 Adadelta 32 Relu 70 0.6714
    下载: 导出CSV

    表 5 

    YOLOV3网络的调参过程

    Table 5. 

    Parameters adjustment of the YOLOV3 model

    学习率 优化器 批次大小 激活函数 轮次 最终的loss
    0.1 Adam 64 Relu 1000 6.107
    0.01 Adam 64 Relu 1000 6.024
    0.001 Adam 64 Relu 1000 2.783
    0.0001 Adam 64 Relu 1000 2.971
    0.001 Adam 64 Relu 1500 2.439
    0.001 Adam 64 Relu 2000 2.131
    0.001 Adam 64 Relu 4000 2.125
    0.001 Adam 128 Relu 2000 2.276
    0.001 Adam 32 Relu 2000 1.701
    0.001 Adam 16 Relu 2000 1.739
    0.001 Adam 32 tanh 2000 4.714
    0.001 Adam 32 Softmax 2000 6.109
    0.001 Adam 32 Sigmoid 2000 5.271
    0.001 RMSprop 32 Relu 2000 3.812
    0.001 Adadelta 32 Relu 2000 3.164
    下载: 导出CSV

    表 6 

    SDNN网络和YOLOV3网络最终的参数设置

    Table 6. 

    Final parameter settings of SDNN network and YOLOV3 network

    网络模型 学习率 批次大小 优化器 激活函数 轮次
    SDNN网络 0.001 32 Adam Relu 70
    YOLOV3网络 0.001 32 Adam Relu 2000
    下载: 导出CSV

    表 7 

    三种闪电哨声波自动检测算法在张衡一号卫星数据中的效果

    Table 7. 

    Results of three different methods for detecting lightning whistler from the data captured by ZH-1 satellite

    错误率 误警率 漏检率 F1score 消耗的时间
    基于形态学处理的闪电哨声波智能检测模型 31% 7% 24% 71.2% 无训练过程
    测试时间:1.21 s/张
    基于SDNN网络的闪电哨声波智能检测模型 17% 2% 15% 81.7% 训练时间:0.16 h;
    测试时间0.39 s/张;
    基于YOLOV3神经网络的闪电哨声波检测模型 0.06% 0.01% 0.05% 98.7% 训练时间:4.5 h;
    测试时间0.12 s/张
    下载: 导出CSV

    表 8 

    闪电哨声波智能检测算法中的哨声波特征提取方法的总结和对比

    Table 8. 

    Summary of the feature extracting methods in the lightning whistler detection algorithms

    特征提取方法 优点 缺点
    人工提取特征(The Stanford VLF Group, 2018; Fiser et al., 2010; Ali Ahmad et al., 2019; Dharma et al., 2014; Oike et al., 2014) 在样本数据少的情况下,人工提取的特征有效,且具有明确的物理含义,可解释性强,时间成本低. 当数据样本量大的时候,人工提取的特征不够鲁棒.
    自动提取特征(Konan et al., 2020) 在样本的数据量足够大的情况能够提取到鲁棒的闪电哨声波的特征. 自动提取的特征的可解释性差;需要大量的训练样本.
    下载: 导出CSV

    表 9 

    闪电哨声波智能检测算法中的哨声波分类(识别)方法的总结和对比

    Table 9. 

    Summary of the classification methods in the lightning whistler detection algorithms

    分类(识别)方法 优点 缺点
    模板匹配分类(The Stanford VLF Group, 2018; Fiser et al., 2010; Ali Ahmad et al., 2019; Dharma et al., 2014; Oike et al., 2014) 简单、快捷,无需训练过程,可直接操作数据,对样本数据量的要求不高. 鲁棒性不强,尤其当测试的数据与模板数据存在差异的时候.
    神经网络分类(Konan et al., 2020) 因为神经网络能够拟合非线性关系,所以其分类效果较佳. 需要模型的训练过程,对训练的机器有硬件方面的要求,容易过拟合.
    下载: 导出CSV

    表 10 

    闪电哨声波智能检测算法中的哨声波定位方法的总结和对比

    Table 10. 

    Summary of the location methods in the lightning whistler detection algorithms

    方法分类 优点 缺点
    滑动窗定位(The Stanford VLF Group, 2018; Fiser et al., 2010; Ali Ahmad et al., 2019; Dharma et al., 2014; Oike et al., 2014; Konan et al., 2020) 简单且效率高,不需要建立数学模型,无需训练过程. 由于定位需要根据阈值进行判断,则定位的精度不够准确.
    智能判断定位(Konan et al., 2020) 定位准确度高 需要建立合理的数学模型,需要训练过程.
    下载: 导出CSV
  •  

    Ali Ahmad U, Kasahara Y, Matsuda S, et al. 2019. Automatic detection of lightning whistlers observed by the plasma wave experiment onboard the arase satellite using the OpenCV library. Remote Sensing, 11(15): 1785. doi: 10.3390/rs11151785

     

    Bayupati I P A, Kasahara Y, Goto Y. 2012. Study of dispersion of lightning whistlers observed by akebono satellite in the Earth's plasmasphere. IEICE Transactions on Communications, 95(11): 3472-3479.

     

    Bradley D, Roth G. 2007. Adaptive thresholding using the integral image. Journal of Graphics Tools, 12(2): 13-21. doi: 10.1080/2151237X.2007.10129236

     

    Bresenham J. 1977. A linear algorithm for incremental digital display of circular arcs. Communications of the ACM, 20(2): 100-106. doi: 10.1145/359423.359432

     

    Brooks C E P. 1925. The distribution of thunderstorms over the globe. Geophys. Mem. London, 1925, 3(24): 147-164.

     

    Cai Z W, Saberian M, Vasconcelos N. 2015. Learning complexity-aware cascades for deep pedestrian detection. //Proceedings of the 2015 IEEE International Conference on Computer Vision. Santiago, Chile: IEEE, 3361-3369.

     

    Chen J T, Inan U S, Bell T F. 1996. VLF strip holographie imaging of lightning-associated ionospheric disturbances. Radio Science, 31(2): 335-348. doi: 10.1029/95RS02079

     

    Clilverd M A, Nunn D, Lev-Tov S J, et al. 2002. Determining the size of lightning-induced electron precipitation patches. Journal of Geophysical Research: Space Physics, 107(A8): SIA 10-1-SIA 10-11. doi: 10.1029/2001JA000301

     

    Dharma K S, Bayupati I P, Buana P W. 2014. Automatic lightning whistler detection using connected component labeling method. Journal of Theoretical and Applied Information Technology, 66(2): 638-645.

     

    Eng H L, Ma K K. 2001. Noise adaptive soft-switching median filter. IEEE Transactions on Image Processing, 10(2): 242-251. doi: 10.1109/83.902289

     

    Fan L L, Zhao H W, Zhao H Y, et al. 2020. Survey of target detection based on deep convolutional neural networks. Optical Precision Engineering (in Chinese), 28(5): 1152-1164.

     

    Felzenszwalb P F, Mcallester D A, Ramanan D. 2008. A discriminatively trained, multiscale, deformable part model//Proceedings of the 2008 IEEE Conference on Computer Vision and Pattern Recognition. Anchorage, Alaska, USA: IEEE, 1-8.

     

    Ferencz C, Ferencz O, Hamar D. 2001. Whistler phenomena: short impulse propagation. Springer Science & Business Media, 160.

     

    Fiser J, Chum J, Diendorfer G, et al. 2010. Whistler intensities above thunderstorms. Annale Geophysicae, 28(1): 37-46. doi: 10.5194/angeo-28-37-2010

     

    Graves A, Mohamed A R, Hinton G. 2013. Speech recognition with deep recurrent neural networks. //Proceedings of the 2013 IEEE International Conference on Acoustics, Speech and Signal Processing. Vancouver, BC, Canada: IEEE, 6645-6649.

     

    Gurnett D A, Kurth W S, Cairns I H, et al. 1990. Whistlers in Neptune's magnetosphere: Evidence of atmospheric lightning. Journal of Geophysical Research: Space Physics, 95(A12): 20967-20976. doi: 10.1029/JA095iA12p20967

     

    Haralick R M, Sternberg S R, Zhuang X H. 1987. Image analysis using mathematical morphology. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-9(4): 532-550. doi: 10.1109/TPAMI.1987.4767941

     

    Helliwell R A. 1965. Whistlers and Related Ionospheric Phenomena. Stanford, Calif. : Stanford University Press.

     

    Horne R B, Glauert S A, Meredith N P, et al. 2013. Space weather impacts on satellites and forecasting the Earth's electron radiation belts with SPACECAST. Space Weather, 11(4): 169-186. doi: 10.1002/swe.20023

     

    Hu Y P, Zeren Z M, Huang J P, et al. 2020. Algorithms and implementation of wave vector analysis tool for the electromagnetic waves recorded by the CSES satellite. Chinese Journal of Geophysics (in Chinese), 63(5): 1751-1765, doi:10.6038/cjg2020N0405.

     

    Huang J P, Lei J G, Li S X, et al. 2018a. The Electric Field Detector (EFD) onboard the ZH-1 satellite and first observational results. Earth and Planetary Physics, 2(6): 469-478, doi:10.26464/epp2018045.

     

    Huang J P, Shen X H, Zhang X M, et al. 2018b. Application system and data description of the China Seismo-Electromagnetic Satellite. Earth and Planetary Physics, 2(6): 444-454, doi:10.26464/epp2018042

     

    Inan U S, Bell T F, Rodriguez J V. 1991. Heating and ionization of the lower ionosphere by lightning. Geophysical Research Letters, 18(4): 705-708. doi: 10.1029/91GL00364

     

    Johnson M P. 2000. VLF imaging of lightning-induced ionospheric disturbances[Ph. D. thesis]. Stanford: Stanford University.

     

    Kingma D P, Ba J. Adam: A method for stochastic optimization. arXiv preprint arXiv: 1412.6980, 2014.

     

    Konan O J E Y, Mishra A K, Lotz S. 2020. Machine learning techniques to detect and characterise whistler radio waves. arXiv preprint arXiv: 2002.01244v1.

     

    Kucherov K I, Nikolaenko A P. 1979. Heating of electrons in the lower ionosphere by horizontal lightning discharges. Radiophysics and Quantum Electronics, 22(7): 621-623. doi: 10.1007/BF01033573

     

    Lay E H, Shao X M. 2011. High temporal and spatial-resolution detection of D-layer fluctuations by using time-domain lightning waveforms. Journal of Geophysical Research: Space Physics, 116(A1): A01317, doi:10.1029/2010JA016018.

     

    LeCun Y, Bengio Y, Hinton G. 2015. Deep learning. Nature, 521(7553): 436-444. doi: 10.1038/nature14539

     

    Lichtenberger J, Ferencz C, Bodnár L, et al. 2008. Automatic whistler detector and analyzer system: Automatic whistler detector. Journal of Geophysical Research: Space Physics, 113(A12): A12201, doi:10.1029/2008JA013467.

     

    Liu S T, Huang D, Wang Y H. 2018. Receptive field block net for accurate and fast object detection. //Proceedings of the 2018 European Conference on Computer Vision (ECCV). Cham: Springer, 385-400.

     

    Lowe D G. 2004. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60(2): 91-110. doi: 10.1023/B:VISI.0000029664.99615.94

     

    Luo X D, Niu S L, Zuo Y H, et al. 2017. Diffusing loss effects of inner radiation belt energetic electrons based on AKEBONO whistler wave parameters. Chinese Journal of Computational Physics (in Chinese), 34(3): 335-343.

     

    Manohar M, Ramapriyan H K. 1989. Connected component labeling of binary images on a mesh connected massively parallel processor. Computer Vision, Graphics, and Image Processing, 45(2): 133-149. doi: 10.1016/0734-189X(89)90129-1

     

    Marshall R, Newsome R, Inan U. 2008b. Fast photometric imaging using orthogonal linear arrays. IEEE Transactions on Geoscience and Remote Sensing, 46(11): 3885-3893. doi: 10.1109/TGRS.2008.2000824

     

    Marshall R A, Inan U S, Chevalier T W. 2008a. Early VLF perturbations caused by lightning EMP-driven dissociative attachment. Geophysical Research Letters, 35(21): L21807, doi:10.1029/2008GL035358.

     

    Masters D, Luschi C. 2018. Revisiting small batch training for deep neural networks. arXiv preprint arXiv: 1804.07612.

     

    Oike Y, Kasahara Y, Goto Y. 2014. Spatial distribution and temporal variations of occurrence frequency of lightning whistlers observed by VLF/WBA onboard Akebono. Radio Science, 49(9): 753-764. doi: 10.1002/2014RS005523

     

    Otsu N. 1979. A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1): 62-66. doi: 10.1109/TSMC.1979.4310076

     

    Parrot M, Achache J, Berthelier J J, et al. 1993. High-frequency seismo-electromagnetic effects. Physics of the Earth and Planetary Interiors, 77(1-2): 65-83. doi: 10.1016/0031-9201(93)90034-7

     

    Parrot M, Berthelier J J, Blecki J, et al. 2015. Unexpected very low frequency (VLF) radio events recorded by the ionospheric satellite DEMETER. Surveys in Geophysics, 36(3): 483-511. doi: 10.1007/s10712-015-9315-5

     

    Parrot M, Pinçon J L, Shklyar D. 2019. Short-fractional hop whistler rate observed by the low-altitude satellite DEMETER at the end of the solar cycle 23. Journal of Geophysical Research: Space Physics, 124(5): 3522-3531. doi: 10.1029/2018JA026176

     

    Pei L S, Ye M, Zhao X Z, et al. 2016. Learning spatio-temporal features for action recognition from the side of the video. Signal, Image and Video Processing, 10(1): 199-206. doi: 10.1007/s11760-014-0726-4

     

    Redmon J, Divvala S, Girshick R, et al. 2016. You only look once: Unified, real-time object detection. //Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA: IEEE, 779-788.

     

    Redmon J, Farhadi A. 2017. YOLO9000: better, faster, stronger. //Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA: IEEE, 7263-7271.

     

    Redmon J, Farhadi A. 2018. YOLOV3: An incremental improvement. arXiv preprint arXiv: 1804. 02767v1.

     

    Santolík O, Parrot M, Inan U S, et al. 2009. Propagation of unducted whistlers from their source lightning: A case study. Journal of Geophysical Research: Space Physics, 114(A3): A03212, doi:10.1029/2008JA013776.

     

    Shen X H, Zhang X M, Cui J, et al. 2018. Remote sensing application in earthquake science research and geophysical fields exploration satellite mission in China. Journal of Remote Sensing (in Chinese), 22(S1): 1993-2002.

     

    The Stanford VLF Group. 2018. Automated Detection of Whistlers for the TARANIS Spacecraft Overview of the Project[EB/OL]. http://vlf.stanford.edu/research/automated-detection-whistlers-taranis-spacecraft.

     

    Viola P, Jones M. 2001. Rapid object detection using a boosted cascade of simple features. //Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Kauai, HI, USA: IEEE, I-1.

     

    Wang Q, Huang J P, Zhang X M, et al. 2018. China Seismo-Electromagnetic Satellite search coil magnetometer data and initial results. Earth and Planetary Physics, 2(6): 462-468, doi:10.26464/epp2018044.

     

    Yang D H, Wang X Y, Shen X H, et al. 2018. Discrimination of shallow and deep induced-microearthquakes based on autoencoder network. Progress in Geophysics (in Chinese), 33(5): 1844-1853, doi:10.6038/pg2018BB0607.

     

    Záhlava J, Němec F, Pincon J L, et al. 2018a. Whistler influence on the overall very low frequency wave intensity in the upper ionosphere. Journal of Geophysical Research: Space Physics, 123(7): 5648-5660. doi: 10.1029/2017JA025137

     

    Záhlava J, Němec F, Santolík O, et al. 2018b. Longitudinal dependence of whistler mode electromagnetic waves in the Earth's inner magnetosphere. Journal of Geophysical Research: Space Physics, 123(8): 6562-6575. doi: 10.1029/2018JA025284

     

    Zeiler M D. 2012. ADADELTA: an adaptive learning rate method. arXiv preprint arXiv: 1212. 5701.

     

    Zhang C X, Zhang J S, Kim S W. 2016. PBoostGA: pseudo-boosting genetic algorithm for variable ranking and selection. Computational Statistics, 31(4): 1237-1262. doi: 10.1007/s00180-016-0652-8

     

    Zhang X M, Qian J D, Shen X H, et al. 2020. The seismic application progress in electromagnetic satellite and future development. Earthquake (in Chinese), 40(2): 18-37.

     

    Zhang Y J, Zhang Y. 2016. The effect research on ionosphere in response to lightning discharge during thunderstorm. Journal of Applied Meteorological Science (in Chinese), 27(5): 570-576.

     

    Zhou Z H. 2016. Machine Learning. Beijing: Tsinghua University Press, 114-115.

     

    胡云鹏, 泽仁志玛, 黄建平等. 2020. 张衡一号卫星记录的空间电磁波传播特征分析方法及算法实现. 地球物理学报, 63(5): 1751-1765, doi:10.6038/cjg2020N0405. http://www.geophy.cn//CN/abstract/abstract15443.shtml

     

    罗旭东, 牛胜利, 左应红等. 2017. 基于AKEBONO哨声波参数的内辐射带高能电子扩散模拟. 计算物理, 34(3): 335-343. https://www.cnki.com.cn/Article/CJFDTOTAL-JSWL201703011.htm

     

    申旭辉, 张学民, 崔静等. 2018. 中国地震遥感应用研究与地球物理场探测卫星计划. 遥感学报, 22(S1): 1993-2002. https://www.cnki.com.cn/Article/CJFDTOTAL-YGXB2018S1001.htm

     

    杨德贺, 王秀英, 申旭辉等. 2018. 基于自编码网络的浅源和深源诱发型微震识别. 地球物理学进展, 33(5): 1844-1853, doi:10.6038/pg2018BB0607.

     

    张学民, 钱家栋, 申旭辉等. 2020. 电磁卫星地震应用进展及未来发展思考. 地震, 40(2): 18-37. https://www.cnki.com.cn/Article/CJFDTOTAL-DIZN202002002.htm

     

    张义军, 张阳. 2016. 雷暴闪电放电活动对电离层影响的研究进展. 应用气象学报, 27(5): 570-576. https://www.cnki.com.cn/Article/CJFDTOTAL-YYQX201605006.htm

     

    周志华. 2016. 机器学习. 北京: 清华大学出版社, 114-115.

  • 加载中

(39)

(10)

计量
  • 文章访问数:  656
  • PDF下载数:  475
  • 施引文献:  0
出版历程
收稿日期:  2020-07-10
修回日期:  2020-12-07
上线日期:  2021-05-10

目录