地球物理学进展 ›› 2020, Vol. 35 ›› Issue (2): 415-421.doi: 10.6038/pg2020CC0295

• 固体地球物理及空间物理学 • 上一篇    下一篇

频高图F层异常分层现象的自动识别

胡辉, 姜春华*(), 周晨, 杨国斌, 赵正予   

  1. 武汉大学电子信息学院,武汉 430072
  • 收稿日期:2019-06-12 修回日期:2019-11-14 出版日期:2020-04-20 发布日期:2020-04-30
  • 通讯作者: 姜春华 E-mail:chuajiang@whu.edu.cn
  • 作者简介:胡辉,男,1993年生,湖北天门人,硕士研究生,研究方向电离层物理与电波传播和空间信号处理. E-mail: 1301112075@qq.com
  • 基金资助:
    国家自然科学基金资助项目(41604133);博士后基金(2016M592374)

Automatic identification of the F-layer stratification on ionograms

HU Hui, JIANG Chun-hua*(), ZHOU Chen, YANG Guo-bin, ZHAO Zheng-yu   

  1. Electronic Information School, Wuhan University, Wuhan 430072, China
  • Received:2019-06-12 Revised:2019-11-14 Online:2020-04-20 Published:2020-04-30
  • Contact: JIANG Chun-hua E-mail:chuajiang@whu.edu.cn

摘要:

频高图F层异常分层(增层)是指在常规电离层结构的基础上,由于电子密度被扰动而在频高图中出现多层回波结构,该F层异常分层(增层)现象在白天和晚上均可观测到.电离层异常分层(增层)主要由某种电动力学或者突发电离源导致电子密度出现扰动而形成,已经引起国内外学者的广泛关注和研究.目前从频高图中识别F层异常分层(增层)主要还是通过人工的方法,海量的频高图数据使该方法在研究F层异常分层(增层)统计特征时变得非常困难.有别于传统的图像分类算法中需要对图像特征进行提取,本文通过引入卷积神经网络(Convolution Neural Network, CNN),设计开发出一种能够自动识别存在F层异常分层(增层)频高图的方法,该方法可以直接输入频高图样本数据,省去了复杂的图像特征提取过程.本文设计的卷积神经网络经过训练以后F层异常分层(增层)识别率为85.82%,准确率90.36%.实验结果表明该卷积神经网络可以较好地自动识别出F层异常分层(增层)的频高图.

关键词: 电离层, 频高图, F层异常分层, 卷积神经网络

Abstract:

In addition to normal layers in the ionosphere, the F-layer stratification (additional layer) is often observed on ionograms due to the disturbances in the electron density. The F-layer stratification (additional layer) could occur both during the daytime and nighttime. The present of disturbances in the ionosphere are mostly caused by the electrodynamics or abnormal ionization source. Currently, the F-layer stratification (additional layer) has attracted extensive attention by many researchers. At present, the identification of the F-layer stratification (additional layer) from ionograms is mainly through the manual method. It is time-consuming task to identify F-layer stratification (additional layer) from a large amount of ionograms. Thus, we propose a new method to automatically identify the F-layer stratification (additional layer) from ionograms using the Convolution Neural Network (CNN). Compared with the image processing algorithms, the present method doesn’t need to preprocess original ionograms to extract some features. Ionograms will be used directly as the input of the method to determine whether the F-layer stratification (additional layer) is present or not. Results show that the recognition rate of F-layer stratification (additional layer) is 85.82% and the accuracy of that is 90.36% through the designed convolution neural network. Moreover, the experimental result indicates that it is effective to automatically identify F-layer stratification (additional layer) on ionograms using the convolution neural network proposed in this paper.

Key words: Ionosphere, Ionogram, F-layer stratification, Convolution Neural Network (CNN)

中图分类号: