基于机器学习的日冕仪图像分类方法研究

日冕物质抛射(Coronal Mass Ejection, CME)的检测是建立CME事件库和实现对CME在行星际传播的预报的重要前提. 通过Visual Geometry Group (VGG) 16卷积神经网络方法对日冕仪图像进行自动分类. 基于大角度光谱日冕仪(Large Angle and Spectrometric Coronagraph Experiment, LASCO) C2的白光日冕仪图像, 根据是否观测到CME对图像进行标记. 将标记分类的数据集用于VGG模型的训练, 该模型在测试集分类的准确率达到92.5%. 根据检测得到的标签结果, 结合时空连续性规则, 消除了误判区域, 有效分类出CME图像序列. 与Coordinated Data Analysis Workshops (CDAW)人工事件库比较, 分类出的CME图像序列能够较完整地包含CME事件, 且对弱CME结构有较高的检测灵敏度. 未来先进天基太阳天文台(Advanced Space-based Solar Observatory, ASO-S)卫星的莱曼阿尔法太阳望远镜将搭载有白光日冕仪(Solar Corona Imager, SCI), 使用此分类方法将该仪器产生的日冕图像按有无CME分类. 含CME标签的图像将推送给中国的各空间天气预报中心, 对CME进行预警.     

基于空时显著性的日冕物质抛射检测

日冕物质抛射(Coronal Mass Ejection, CME)是一种剧烈的太阳爆发现象, 它会对行星际空间造成严重扰动, 进而影响人类生产、生活. 基于CME的时空显著性, 将显著性检测方法引入到CME检测中, 利用结构化矩阵分解SOHO (Solar and Heliospheric Observatory)的大角度光谱日冕仪(Large Angle and Spectrometric Coronagraph Experiment, LASCO) C2的日冕图像对应的特征矩阵, 从中恢复出稀疏部分获得显著前景. 然后考虑CME运动时产生的时间显著性, 从而去除非CME结构(如冕流), 得到最终检测结果. 实验表明, 以人工目录协调数据分析中心(Coordinated Data Analysis Workshop, CDAW)检测结果为基准时, 所提方法不仅在检测CME数量上比计算机辅助跟踪软件包(Computer Aided CME Tracking Software package, CACTus)和太阳爆发事件检测系统(Solar Eruptive Event Detection System, SEEDS)有优势, 还在CME中心角度和张角宽度等特征物理参数测量上比CACTus和SEEDS更接近CDAW目录参考值.     

对日冕物质抛射分类的定量研究

众所周知，日冕物质抛射（CME）的性质强烈地决定于其结构，故对日冕物质抛射的结构进行强度分类，对研究日地事件有重要的意义，利用模糊综合评判方法，可对日冕物质抛射的分类进行定量研究，该方法通常需考虑多个因素，如各类CME的速度、跨度、质量和动能。将其作为评判因素集，并确定相应的隶属度函数，通过计算，可以将CME分成强（strong）、中（middle）、弱（weak）3大类，并根据隶属度值的大小，给出各类CME平均特性的相对强弱排序，分析表明模糊综合评判方法优于一般的统计方法。     

STEREO卫星的CME观测研究进展

简要回顾利用"日地关系天文台"(Solar Terrestrial Relations Observatory,STEREO)卫星的立体观测资料在日冕物质抛射(Coronal Mass Ejection,CME)研究方面已取得的一些重要进展,主要包括(1)通过极紫外成像仪观测到的日冕极紫外暗化来更准确地估计CME质量,研究CME演化的结构特征;(2)利用STEREO卫星日冕仪的双角度观测,在CME立体传播特征方面取得的新进展;(3)STEREO卫星日球成像仪具有广阔的视场范围,可以跟踪研究CME从太阳表面爆发到形成行星际日冕物质抛射(Interplanetary CME,ICME),及其在内日球层和近地空间的演化特征以及运动特征等。同时,也介绍了利用三角测量技术测定CME特征物理量的新方法。     

不同速度日冕物质抛射的多波段射电观测特征

比较了12个日冕物质抛射(CME)事件, 发现它们可以分为两类, 其中分别是快速(＞1000 km/s)和慢速(≤800 km/s)各6个事件, 发现这2类CME事件分别对应于不同的多波段射电辐射类型和不同的日冕磁位形.本文定性地分析了这二个类型的射电爆发的产生过程,指出多重磁极和双磁极结构可能是分别产生二类CME和二类多波段射电爆发类型的原因,并涉及到"磁崩溃"模型与多重磁结构的关系.讨论了CME的不同速度可能是造成多波段不同射电爆发的主要因素,并指出快速或慢速的CME可能取决于日冕的多重或双重磁结构.     

日冕物质抛射及其驱动激波的多波段和多视角研究

<正>日冕物质抛射(Coronal Mass Ejection, CME)是太阳大气中剧烈的爆发现象之一.其爆发通常能释放大量的能量并抛射大量磁化等离子体. CME所驱动的激波能进一步导致太阳高能粒子事件(Solar Energetic Particle,SEP)的发生,并可能影响航天器和宇航员的安全.因此,研究CME及其驱动激波的形成机制和性质有利于我们更加清晰地了解及监测它们的运动过程,     

日冕物质抛射的偏转特性研究

日冕物质抛射(CME)是巨大的、携带磁力线的泡沫状气体,在几个小时中被从太阳抛射出来的过程。日冕物质抛射伴随着大量带电粒子和辐射的释放,这些物质进入日地空间,对日地空间的磁场造成很大扰动;当它们传播到地球附近时,则严重影响地球的磁场,产生磁暴,也对空间和地面的电子设备造成干扰。日冕物质抛射在传播过程中如果发生偏转,将影响它对地有效性。因此研究日冕物质抛射的偏转特性,对预报日冕物质抛射对日地空间的影响具有重要意义。主要利用2007年10月8日STEREO卫星的日冕物质抛射观测资料,结合全日面线性无力场模型(Global Linear Force-Free Field,GLFFF)进行磁场外推,分析日冕物质抛射偏转与背景磁场能量密度分布之间的关系,并计算日冕物质抛射的运动轨迹。通过改变无力因子α,发现当α=0.15时,计算得到的日冕物质抛射运动轨迹与实际观测的日冕物质抛射运动轨迹拟合得最好。     

一个复杂大爆发的分析

本文分析了2002.7.23国家天文台云南天文台射电频谱仪在625～1500 MHz、2600～3800 MHz和5200～7600 MHz记录到的复杂型大爆发,将此爆发与Hα耀斑、日冕物质抛射(CME)、硬X射线爆发及地球物理参数作了相关分析,得到这个事件的一些显著特征,认为这一事件电子的加速区在日冕的外层,接近625 MHz的地方,并且多次发生磁重联.磁重联以后的衰变相是湍流加速过程.     

晕状日冕物质抛射

主要介绍晕状日冕物质抛射（halo CMEs)的产生机制，包括向量磁场演化是怎样触发halo CMEa的：halo CME与耀斑，暗条活动的相互关系怎样，是否有规律可循，暗条爆发，耀斑等活动现象是如何相互联系的，halo CME事件是由一个活动区域或一个活动事件驱动物，还是多个活动区或多个活动事件相互作用的结果，给出两个halo CME的日面起源的观测例证，提出相反极笥的磁场对消是CME日面源区磁场演化的主要特征。     

惊天动地的CME

CME是(Coronal Mass Ejection)的缩写,意为日冕物质抛射。 太阳耀斑爆发已经是一个规模巨大的、剧烈的活动现象了。CME则是太阳日冕层中规模比太阳耀斑还大许多倍的活动现象,或者说是尺度最大、最壮观的爆发现象。从物理意义上讲,CME是从太阳向外喷射出的庞大等离子体和磁场结构,是日冕和太阳风     

Automatic Detection and Classification of Coronal Mass Ejections

We present an automatic algorithm to detect, characterize, and classify coronal mass ejections (CMEs) in Large Angle Spectrometric Coronagraph (LASCO) C2 and C3 images. The algorithm includes three steps: (1) production running difference images of LASCO C2 and C3; (2) characterization of properties of CMEs such as intensity, height, angular width of span, and speed, and (3) classification of strong, median, and weak CMEs on the basis of CME characterization. In this work, image enhancement, segmentation, and morphological methods are used to detect and characterize CME regions. In addition, Support Vector Machine (SVM) classifiers are incorporated with the CME properties to distinguish strong CMEs from other weak CMEs. The real-time CME detection and classification results are recorded in a database to be available to the public. Comparing the two available CME catalogs, SOHO/LASCO and CACTus CME catalogs, we have achieved accurate and fast detection of strong CMEs and most of weak CMEs.     

Detection of Coronal Mass Ejections Using Multiple Features and Space–Time Continuity

Coronal Mass Ejections (CMEs) release tremendous amounts of energy in the solar system, which has an impact on satellites, power facilities and wireless transmission. To effectively detect a CME in Large Angle Spectrometric Coronagraph (LASCO) C2 images, we propose a novel algorithm to locate the suspected CME regions, using the Extreme Learning Machine (ELM) method and taking into account the features of the grayscale and the texture. Furthermore, space–time continuity is used in the detection algorithm to exclude the false CME regions. The algorithm includes three steps: i) define the feature vector which contains textural and grayscale features of a running difference image; ii) design the detection algorithm based on the ELM method according to the feature vector; iii) improve the detection accuracy rate by using the decision rule of the space–time continuum. Experimental results show the efficiency and the superiority of the proposed algorithm in the detection of CMEs compared with other traditional methods. In addition, our algorithm is insensitive to most noise.     

Automatic Detection and Tracking of Coronal Mass Ejections in Coronagraph Time Series

We present the current capabilities of a software tool to automatically detect coronal mass ejections (CMEs) based on time series of coronagraph images: the solar eruptive event detection system (SEEDS). The software developed consists of several modules: preprocessing, detection, tracking, and event cataloging. The detection algorithm is based on a 2D to 1D projection method, where CMEs are assumed to be bright regions moving radially outward as observed in a running-difference time series. The height, velocity, and acceleration of the CME are automatically determined. A threshold-segmentation technique is applied to the individual detections to automatically extract an approximate shape of the CME leading edge. We have applied this method to a 12-month period of continuous coronagraph images sequence taken at a 20-minute cadence by the Large Angle and Spectrometric Coronagraph (LASCO) instrument (using the C2 instrument only) onboard the Solar and Heliospheric Observatory (SOHO) spacecraft. Our automated method, with a high computational efficiency, successfully detected about 75% of the CMEs listed in the CDAW CME catalog, which was created by using human visual inspection. Furthermore, the tool picked up about 100% more small-size or anomalous transient coronagraph events that were ignored by human visual inspection. The output of the software is made available online at . The parameters of scientific importance extracted by the software package are the position angle, angular width, velocity, peak, and average brightness. Other parameters could easily be added if needed. The identification of CMEs is known to be somewhat subjective. As our system is further developed, we expect to make the process significantly more objective.     

Detection of coronal mass ejections using AdaBoost on grayscale statistic features

We present an automatic algorithm to detect coronal mass ejections (CMEs) in Large Angle Spectrometric Coronagraph (LASCO) C2 running difference images. The algorithm includes 3 steps: (1) split the running difference images into blocks according to slice size and analyze the grayscale statistics of the blocks from a set of images with and without CMEs; (2) select the optimal parameters for slice size, gray threshold and fraction of the bright points and (3) use AdaBoost to combine the weak classifiers designed according to the optimal parameters. Experimental results show that our method is effective and has a high accuracy rate.     

Multi-label Learning for Detection of CME-Associated Phenomena

Coronal mass ejections (CMEs) are considered as one of the driving sources of space weather. They are usually associated with many physical phenomena, e.g. flares, coronal dimmings, and sigmoids. To detect these phenomena, traditional supervised-learning methods assumed that at most one event occurred in a CME; therefore each CME instance is associated with a single label and the phenomenon is processed in isolation. This simplifying assumption does not fit well, as CMEs might have multiple events simultaneously. We propose to detect multiple CME-associated events by multi-label learning methods. With the data available from the Atmospheric Imaging Assembly (AIA) and the Large Angle and Spectrometric Coronagraph (LASCO), texture features representing the events are extracted from all of the associated and not-associated CMEs and converted into feature vectors for multi-label learning use. Then a function is learned to predict the proper label sets for CMEs, such that eight events, i.e. coronal dimming, coronal hole, coronal jet, coronal wave, filament, filament eruption, flare, and sigmoid, are detected explicitly. To test the proposed detection algorithm, we adopt the four-fold cross-validation strategy on a set of 551 labeled CMEs from AIA. Experimental results demonstrate the good performance of the multi-label classification methods in terms of test error.     

Speed Distributions of CMEs in Cycle 23 at Low and High Latitudes

We analyzed the speed (v) distributions of 11584 coronal mass ejections (CMEs) observed by the Large Angle and Spectrometric Coronagraph Experiment on board the Solar and Heliospheric Observatory (SOHO/LASCO) in cycle 23 from 1996 to 2006. We find that the speed distributions for high-latitude (HL) and low-latitude (LL) CME events are nearly identical and to a good approximation they can be fitted with a lognormal distribution. This finding implies that statistically the same driving mechanism of a nonlinear nature is acting in both HL and LL CME events, and CMEs are intrinsically associated with the source's magnetic structure on large spatial scales. Statistically, the HL CMEs are slightly slower than the LL CMEs. For HL and LL CME events respectively, the speed distributions for accelerating and decelerating events are nearly identical and also to a good approximation they can be both fitted with a lognormal distribution, thus supplementing the results obtained by Yurchyshyn et al.     

Velocity Distribution of CMEs After Projection Correction

The observed CME (coronal mass ejection) is its projection on the sky plane, and this leads to certain discrepancies between the observational and true parameters of the CME. For example, the observed velocity is generally smaller than the true velocity. The method of making projection correction for the CME velocity based on the conical model is utilized to analyze the velocity distributions of the 1691 CMEs which are only correlated to flares (called the class FL CMEs for short) and the 610 CMEs which are only correlated to filament eruptions (called the class FE CMEs for short) before and after the projection correction. These CMEs were observed with the Large Angle and Spectrometric Coronograph on the Solar and Heliospheric Observatory from September 1996 to September 2007 (close to a solar cycle). The obtained results are as follows: (1) before and after the projection correction the velocity distribution of FL CMEs is quite similar to that of FE CMEs, and before and after the projection correction the mean velocities of the two classes of CMEs are almost the same; (2) before and after the projection correction, the natural logarithm distribution of the FL CME velocities is also very similar to that of the FE CME velocities.