极低频电磁台网成功观测到大量的Pc1地磁脉动事件,研究极低频Pc1地磁脉动的自动识别方法对于全面分析地球空间电磁物理环境具有重要意义.本文采用了 YOLOv8目标检测网络、ResNet残差网络和定向特征增强技术,提出了一种基于计算机视觉的Pc1地磁脉动自动识别模型(Automatic Detection Model for Pc1 Geomagnetic Pulsation,简称ADM-Pc1).以大连台站和丽江台站的极低频观测数据为例,利用2015-2016年的数据作为训练集进行模型的监督学习,并使用2017-2022年的数据作为测试集对模型性能进行评估.实验结果显示,ADM-Pc1模型的F1-Score值达到了 95％,错分率仅为0.9％,虚警率仅为5.8％,漏检率仅为9％,处理1天数据平均耗时是2.72 s,显著优于现有的最优识别模型.这表明,ADM-Pc1模型在识别效果和计算速度方面均能更好地满足实际工程需求.
Automatic detection model for Pc1 geomagnetic pulsation in Extremely Low Frequency data
The Extremely Low Frequency(ELF)electromagnetic network has successfully detected numerous Pc1 geomagnetic pulsation events.To facilitate comprehensive analysis of the geospace environment,it is essential to automatically identify Pc1 events from massive observational data.This study proposes an automated detection model named the Automatic Detection Model for Pc1 Geomagnetic Pulsation(ADM-Pc1).The ADM-Pc1 utilizes the YOLOv8 object detection network,ResNet residual network,and directional feature enhancement technology.This study utilized low-frequency observation data from Dalian and Lijiang electromagnetic stations as a case study.Data from 2015 to 2016 was employed as the training set for supervised learning of the model,while data from 2017 to 2022 served as the test set for evaluating the model's performance.The ADM-Pc1 achieved an F1-score of 95％,a misclassification rate of 0.9％,a false alarm rate of 5.8％,a missed detection rate of 9％,and an average processing time of 2.72 seconds per day of data,outperforming most state-of-the-art methods.These results indicate that the ADM-Pc1 model exhibits high accuracy and robustness.Future work will focus on further developing the automatic recognition prototype system for Pc1 geomagnetic pulsations to meet practical engineering requirements.
Pc1 geomagnetic pulsationsComputer visionAutomatic identificationExtremely Low Frequency electromagnetic network
万方数据
维普
