Pattern Recognition and Leading Edge Extraction for the Backscatter Ionograms by YOLOX  

基于YOLOX对斜返电离图的模式识别和前沿提取

作　　者：MA Jiasheng YANG Guobin JIANG Chunhua LIU Tongxin 马家晟;杨国斌;姜春华;刘桐辛(武汉大学电子信息学院,湖北武汉430072)

机构地区：[1]Electronic Information School,Wuhan University,Wuhan 430072,Hubei,China

出　　处：《Wuhan University Journal of Natural Sciences》2025年第1期69-78,共10页武汉大学学报(自然科学英文版)

基　　金：Supported by the National Natural Science Foundation of China(42104151,42074184,42188101,41727804)。

摘　　要：The ionosphere is an important component of the near Earth space environment.The three common methods for detecting the ionosphere with high frequency(HF)radio signals are vertical detection,oblique detection,and oblique backscatter detection.The ionograms obtained by these detection methods can effectively reflect a large amount of effective information in the ionosphere.The focus of this article is on the oblique backscatter ionogram obtained by oblique backscatter detection.By extracting the leading edge of the oblique backscatter ionogram,effective information in the ionosphere can be inverted.The key issue is how to accurately obtain the leading edge of the oblique backscatter ionogram.In recent years,the application of pattern recognition has become increasingly widespread,and the YOLO model is one of the best fast object detection algorithms in one-stage.Therefore,the core idea of this article is to use the newer YOLOX object detection algorithm in the YOLO family to perform pattern recognition on the F and E_(s) layers echoes in the oblique backscatter ionogram.After image processing,a single-layer oblique backscatter echoes are obtained.It can be found that the leading edge extraction of the oblique backscatter ionogram obtained after pattern recognition and image processing by the YOLOX model is more fitting to the actual oblique backscatter leading edge.电离层是近地空间环境的一个重要组成部分。利用高频(HF)无线电信号探测电离层的三种常见方法是垂直探测、斜向探测和斜向返回散射探测。通过这些探测方法获得的电离图可以有效地反映电离层中的大量有效信息。本文的研究重点是通过斜向返回散射探测获得的斜向返回散射电离图。通过提取斜向返回散射电离图的前沿,可以反演电离层中的有效信息。如何准确地获得斜向返回散射电离图的前沿是本文研究的关键问题。近年来,模式识别的应用越来越广泛,YOLO模型是单阶段快速目标检测的最佳算法之一。因此,本文的核心思想是使用YOLO家族中较新的YOLOX目标检测算法对斜向返回散射电离图中的F和E_(s)层回波进行模式识别。将模式识别后的斜向返回散射电离图经过图像处理后,得到单层斜向返回散射回波的电离图。可以发现,通过YOLOX模型进行模式识别和图像处理后获得的斜向返回散射电离图的前沿描迹更符合实际情况下的斜向返回散射前沿描迹。

关 键 词：IONOSPHERE backscatter ionogram leading edge YOLOX pattern recognition 

分 类 号：P352.7[天文地球—空间物理学]