基于深度学习的频高图自动标定  

Ionogram automatic scaling based on Deep-learning

作　　者：朱正平[1] 邢蕴辉 ZHU Zhengping;XING Yunhui(College of Electronics and Information Engineering,South-Central Minzu University,Wuhan 430074,China)

机构地区：[1]中南民族大学电子信息工程学院,武汉430074

出　　处：《中南民族大学学报(自然科学版)》2025年第4期497-506,共10页Journal of South-Central Minzu University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(41474135)。

摘　　要：频高图是电离层测高仪地面探测电离层的常规数据,数据量大,电离层各种参数是需要进行逐一标定才能获取,传统上需要手工标定,费时费力且易出错,实现计算机辅助下的手动标定势在必行.为此提出了一种基于深度学习的频高图自动标定方法,基于U型结构并采用具有横向连接的特征金字塔作为连接结构,参照频高图手动标定数据,生成网络模型样本数据,然后随机选取部分数据作为训练数据输入,通过不断更新参数,使得网络模型的预测值逐渐接近真实值.结果显示,与自动标定程序ARTIST相比,文中所提出的模型在精度和召回率方面分别提高了8%和17%,且自动标定结果与手动标定结果相近.这表明基于深度学习方法自动标定的频高图可应用于全球电离层天气实时预报.Ionogram is the conventional data of ionospheric sounding by ionosonde on the ground.The amount of data is large.Various parameters of the ionosphere need to be scaled one by one to obtain.Traditionally,manual scaling is required but it is time-consuming,laborious and error-prone.It is imperative to realize computer-assisted manual scaling potential.Herein a deep-learning method for ionogram automatic scaling(DIAS)is presented,and the method is based on the U-shaped structure and using the characteristic pyramid with horizontal connection as the connection,using the data of manual scaling to generate the model sample data,and then randomly select part of the data as the training data input,so that the predicted value of the model gradually approaches the true value by constantly updating the parameters.The results show that compared with Automatic Real-Time Ionogram Scaling with True-height(ARTIST),the accuracy and recall rate of DIAS are improved by 8%and 17%respectively,and the results of DIAS are similar to those of manual scaling.This results shows that ionograms provided by deep-learning method can be applied to real-time global ionospheric weather nowcasting.

关 键 词：频高图标定 深度学习 电离层测高仪 电离层 

分 类 号：P352[天文地球—空间物理学]