袁静^1,,
王桥^2,,,
杨德贺²,
刘芹芹²,
泽仁志玛²,
申旭辉²
1. 防灾科技学院, 河北三河 065421
2. 应急管理部国家自然灾害防治研究院, 北京 100085

基金项目: 国家重点研发计划（2018YFC1503501，2018YFC1503806），中央直属高校基本科研业务（ZY20180122），亚太空间合作组织地震二期项目（APSCO Earthquake Research Project Phase Ⅱ）和国际空间科学研究所北京分部项目（ISSI-BJ，2019IT-33），廊坊科技局科学研究与发展计划（2020011025）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: 王桥, 男, 1986年出生, 助理研究员, 博士, 2015年毕业于北京大学, 研究方向为地球电磁学, 目前负责张衡一号感应式磁力仪数据处理与应用研究.E-mail: qiaowang@ninhm.ac.cn

中图分类号: P352

收稿日期:2020-05-07
修回日期:2021-01-29
上线日期:2021-11-10

Automatic recognition algorithm of lightning whistlers observed by the Search Coil Magnetometer onboard the Zhangheng-1 Satellite

YUAN Jing^1,,
WANG Qiao^2,,,
YANG DeHe²,
LIU QinQin²,
ZHIMA ZeRen²,
SHEN XuHui²
1. Institute of Disaster Prevention, Sanhe Hebei 065421, China
2. National Institute of Natural Hazards, Ministry of Emergency Management of China, Beijing 100085, China


More Information

Corresponding author: WANG Qiao,E-mail:qiaowang@ninhm.ac.cn

MSC: P352

--> Received Date: 07 May 2020
Revised Date: 29 January 2021
Available Online: 10 November 2021


摘要
摘要:张衡一号卫星感应式磁力仪（Search Coil Magnetometer，SCM）探测到了大量的低频电磁波动数据.本文探索从中自动识别闪电哨声波（Lightning Whistler，LW）的算法，相关结果对进一步研究空间天气闪电事件的时空变化规律具有重要研究意义.首先，以20 s的时间窗提取SCM原始波形数据，再对其做短时傅里叶变换（Short-Time Fourier Transform，STFT）得到时频图；接着，以LW在时频图中呈现明显的L形态特征为依据创建LW时频图像数据集，该数据集包括316个LW时频图，8000个非闪电哨声波的时频图；再对时频图像进行灰度化处理和尺度缩放处理以降低计算维度，同时增强闪电哨声波特征；通过设计模糊卷积核，对图像进行卷积计算以滤除大量阶跃边缘信息的影响；基于LW的形态特征设计L形态卷积核，对图像进行卷积处理以进一步增强图像中的L形态特征.最后，将增强后的L特征图输入支持向量机（Support Vector Machines，SVM）进行分类识别.大量数据处理实验结果表明：本文提出的闪电哨声波自动识别方案有效，其识别效果在精度、召回率、F1值（F1 score）和接受者操作特性曲线面积（Area Under Curve，AUC）指标上均达到94%以上.
关键词: 张衡卫星/
感应磁力仪/
闪电哨声波/
识别

Abstract:Zhangheng-1 satellite has been recording a large number of electromagnetic fluctuations from the search coil magnetometer (SCM). How to automatically recognize lightning whistlers from the data is important to further explore the temporal and spatial variation of lightning events. Firstly, SCM wave data is processed by Short Time Fourier Transformation (STFT) to obtain the Fourier spectrogram. When the lightning whistlers occur, the L-shape could be found in the spectrogram, hence, the spectrogram image was segmented to obtain data group including 316 sub-images with lightning whistlers and 8000 ones without lightning whistlers; secondly, all the sub-images in the data group should be processed by image processing techniques to enhance the lightning whistlers; thirdly, the fuzzy convolution kernel is proposed to process the sub-images to filter out the influence of a large number of step edge information. Next, the L-shape convolution kernel is proposed to further enhance the L-shape feature in the image. Finally, the enhanced images as feature vectors are input into the support vector machine (SVM) to train the recognition model. The experimental results show that the proposed automatic lightning whistlers recognition algorithm is effective, and it reaches over 94% both in accuracy, recall rate, F1 value (F1 score), and area under curve (AUC) index.
Key words:Zhangheng-1 satellite/
Search coil magnetometer/
Lightning whistler/
Recognition



PDF全文下载地址:

http://www.geophy.cn/data/article/export-pdf?id=8b9db41f-ec06-4adc-a2e1-32b8a8623bb3