基于深度学习的太阳活动区检测与跟踪方法研究  

A Detection and Tracking Method for Active Regions Based on Deep Learning

作　　者：朱健 杨云飞[1,2] 苏江涛 刘海燕 李小洁[1] 梁波 冯松 Zhu Jian;Yang Yunfei;Su Jiangtao;Liu Haiyan;Li Xiaojie;Liang Bo;Feng Song(School of Information Engineering and Automation/Key Laboratory of Applications of Computer Technology of the Yunnan Province,Kunming University of Science and Technology,Kunming 650500,China;Key Laboratory of Solar Activity,Chinese Academy of Sciences,Beijing 100101,China)

机构地区：[1]昆明理工大学信息工程与自动化学院/云南省计算机技术应用重点实验室,云南昆明650500 [2]中国科学院太阳活动重点实验室,北京100101

出　　处：《天文研究与技术》2020年第2期191-200,共10页Astronomical Research & Technology

基　　金：国家自然科学基金(11763004,11573012,11803085,U1931107);中国科学院太阳活动重点实验室开放课题(KLSA202019);中国科学院空间科学战略性先导科技专项(XDA15320302,XDA15052200,XDA15320102);云南省重点研发计划(2018IA054)资助.

摘　　要：太阳活动区是各类太阳活动的主要能量来源,剧烈的太阳活动直接影响人类的生存环境,因此,准确地检测与跟踪太阳活动区对监控和预报空间天气非常重要.基于深度学习框架的YOLOv3-spp和DeepSort,提出了一种太阳活动区检测和跟踪方法(Active Regions Detection and Tracking Method,ARDTM),该方法较好地解决了传统图像处理方法易将一个太阳活动区误检测为多个,或者多个太阳活动区误检测为一个的问题;及时捕获新产生的太阳活动区和终止跟踪消失的太阳活动区,有效提高了太阳活动区的跟踪准确率.实验结果表明,该方法可以较好地检测和跟踪不同望远镜、不同时间间隔序列图像中的太阳活动区.Active regions(ARs)are the primary source of energy for various solar activities.The violent solar activities adversely affect human living environment.Therefore,accurate detection and tracking of ARs are very important for monitoring and forecasting the space weather.In this paper,we propose an AR detection and tracking method(ARDTM)based on the deep learning model comprising of the YOLOv3-spp and DeepSort.The method solves the problem that one AR is mis-detected as multiple ARs,or multiple ARs are mis-detected as an AR.Besides,it captures new ARs and terminates disappeared ARs in time.The method improves the precision of detecting and tracking ARs.It can be used for detecting and tracking ARs in the solar full-disk longitudinal magnetograms observed from different telescopes,or images of different time interval series.

关 键 词：太阳活动区 检测与跟踪 深度学习 YOLOv3-spp DeepSort 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]