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.     

CME速度的投影效应改正方法

目前观测得到的日冕物质抛射(coronal mass ejection,CME)只是其在天空平面的投影,其观测参量与真实参量之间存在一定的差异.而CME的速度是对其地磁效应有决定性影响的参量,因此对CME测量速度作投影效应改正是一个重要的研究课题.综述了近年来对CME测量速度进行投影效应改正的方法,并指出了这些投影效应改正方法中存在的一些问题和进一步的研究方向.     

Tracking of Coronal White-Light Events by Texture

The extraction of the kinematic properties of coronal mass ejections (CMEs) from white-light coronagraph images involves a significant degree of user interaction: defining the edge of the event, separating the core from the front or from nearby unrelated structures, etc. To contribute towards a less subjective and more quantitative definition, and therefore better kinematic characterization of such events, we have developed a novel image-processing technique based on the concept of “texture of the event”. The texture is defined by the so-called gray-level co-occurrence matrix, and the technique consists of a supervised segmentation algorithm to isolate a particular region of interest based upon its similarity with a pre-specified model. Once the event is visually defined early in its evolution, it is possible to automatically track the event by applying the segmentation algorithm to the corresponding time series of coronagraph images. In this paper we describe the technique, present some examples, and show how the coronal background, the core of the event, and even the associated shock (if one exists) can be identified for different kind of CMEs detected by the LASCO and SECCHI coronagraphs.     

Validation of CME Detection Software (CACTus) by Means of Simulated Data,and Analysis of Projection Effects on CME Velocity Measurements

In the context of space weather forecasting, an automated detection of coronal mass ejections (CMEs) becomes more and more important for efficiently handling a large data flow which is expected from recently-launched and future solar missions. In this paper we validate the detection software package “CACTus” by applying the program to synthetic data from our 3D time-dependent CME simulations instead of observational data. The main strength of this study is that we know in advance what should be detected. We describe the sensitivities and strengths of automated detection, more specific for the CACTus program, resulting in a better understanding of CME detection on one hand and the calibration of the CACTus software on the other hand, suggesting possible improvements of the package. In addition, the simulation is an ideal tool to investigate projection effects on CME velocity measurements.     

Estimating Travel Times of Coronal Mass Ejections to 1 AU Using Multi-spacecraft Coronagraph Data

We study the relationship between the speeds of coronal mass ejections (CMEs) obtained close to the Sun and in the interplanetary medium during the low solar-activity period from 2008 to 2010. We use a multi-spacecraft forward-modeling technique to fit a flux-rope-like model to white-light coronagraph images from the STEREO and SOHO spacecraft to estimate the geometrical configuration, propagation in three-dimensions (3D), and the radial speeds of the observed CMEs. The 3D speeds obtained in this way are used in existing CME travel-time prediction models. The results are compared to the actual CME transit times from the Sun to STEREO, ACE, and Wind spacecraft as well as to the transit times calculated using projected CME speeds. CME 3D speeds give slightly better predictions than projected CME speeds, but a large scatter is observed between the predicted and observed travel times, even when 3D speeds are used. We estimate the possible sources of errors and find a weak tendency for large interplanetary CMEs (ICMEs) with high magnetic fields to arrive faster than predicted and small, low-magnetic-field ICMEs to arrive later than predicted. The observed CME transit times from the Sun to 1?AU show a particularly good correlation with the upstream solar-wind speed. Similar trends have not been observed in previous studies using data sets near solar maximum. We suggest that near solar minimum a relatively narrow range of CME initial speeds, sizes, and magnetic-field magnitudes led to a situation where aerodynamic drag between CMEs and ambient solar wind was the primary cause of variations in CME arrival times from the Sun to 1?AU.     

Solar Sources of Interplanetary Coronal Mass Ejections During the Solar Cycle 23/24 Minimum

We examine solar sources for 20 interplanetary coronal mass ejections (ICMEs) observed in 2009 in the near-Earth solar wind. We performed a detailed analysis of coronagraph and extreme ultraviolet (EUV) observations from the Solar Terrestrial Relations Observatory (STEREO) and Solar and Heliospheric Observatory (SOHO). Our study shows that the coronagraph observations from viewpoints away from the Sun–Earth line are paramount to locate the solar sources of Earth-bound ICMEs during solar minimum. SOHO/LASCO detected only six CMEs in our sample, and only one of these CMEs was wider than 120^°. This demonstrates that observing a full or partial halo CME is not necessary to observe the ICME arrival. Although the two STEREO spacecraft had the best possible configuration for observing Earth-bound CMEs in 2009, we failed to find the associated CME for four ICMEs, and identifying the correct CME was not straightforward even for some clear ICMEs. Ten out of 16 (63 %) of the associated CMEs in our study were “stealth” CMEs, i.e. no obvious EUV on-disk activity was associated with them. Most of our stealth CMEs also lacked on-limb EUV signatures. We found that stealth CMEs generally lack the leading bright front in coronagraph images. This is in accordance with previous studies that argued that stealth CMEs form more slowly and at higher coronal altitudes than non-stealth CMEs. We suggest that at solar minimum the slow-rising CMEs do not draw enough coronal plasma around them. These CMEs are hence difficult to discern in the coronagraphic data, even when viewed close to the plane of the sky. The weak ICMEs in our study were related to both intrinsically narrow CMEs and the non-central encounters of larger CMEs. We also demonstrate that narrow CMEs (angular widths ≤?20^°) can arrive at Earth and that an unstructured CME may result in a flux rope-type ICME.     

Three-Dimensional Kinematics of Coronal Mass Ejections from STEREO/SECCHI-COR2 Observations in 2007 – 2008

We present a new method to perform the three dimensional characterization of coronal mass ejections (CMEs) using stereoscopic images obtained with the STEREO/SECCHI-COR2 coronagraphs. Two approaches are proposed, and each associated algorithm gives the trajectory of the CME and its kinematical properties (velocity and acceleration profiles) intended for space weather forecast. The first approach is based on forward modeling appropriate to the reconstruction of surfaces in an optically thin medium, and performs a local approximation of the observed CME by a hemispherical shell, thus tracking the leading edge of the event. The second approach is based on tracking the center of gravity of the radiance of the CME in the images. More than 16?000 blind tests have been performed to assess the performance of each algorithm. For that purpose, we used three distinct libraries of simulated images of CMEs that correspond to three CME models: hemispherical shell, flux rope, and cloud-like. The two methods are applied to a set of CMEs observed in 2007 and 2008 by the SECCHI-COR2 coronagraphs, and when possible, our results are compared to those already published. The determinations of the direction of propagation and of the velocity are generally found in good agreement.     

Onboard Automated CME Detection Algorithm for the Visible Emission Line Coronagraph on <Emphasis Type="Italic">ADITYA-L1</Emphasis>

ADITYA-L1 is India’s first space mission to study the Sun from the Lagrange 1 position. The Visible Emission Line Coronagraph (VELC) is one of seven payloads on the ADITYA-L1 mission, which is scheduled to be launched around 2020. One of the primary objectives of the VELC is to study the dynamics of coronal mass ejections (CMEs) in the inner corona. This will be accomplished by taking high-resolution (\({\approx}\,2.51~\mbox{arcsec}\,\mbox{pixel}^{-1}\)) images of the corona from \(1.05~\mbox{R}_{\odot}\,\mbox{--}\,3~\mbox{R}_{\odot}\) at a high cadence of 1 s in the 10 Å passband centered at 5000 Å. Because telemetry at the Lagrangian 1 position is limited, we plan to implement an onboard automated CME detection algorithm. The detection algorithm is based on intensity thresholding followed by area thresholding in successive difference images that are spatially rebinned to improve the signal-to-noise ratio. We present the results of the application of this algorithm on the data from existing coronagraphs such as STEREO/SECCHI COR-1, which is a space-based coronagraph, and K-Cor, a ground-based coronagraph, because they have a field of view (FOV) that is most similar to that of VELC. Since no existing space-based coronagraph has a FOV similar to VELC, we have created synthetic coronal images for the VELC FOV after including photon noise and injected CMEs of different types. The performance of the CME detection algorithm was tested on these images. We found that for VELC images, the telemetry can be reduced by a factor of 85% or more while maintaining a CME detection rate of 70% or higher at the same time. Finally, we discuss the advantages and disadvantages of this algorithm. The application of such an onboard algorithm in future will enable us to take higher resolution images with an improved cadence from space and simultaneously reduce the load on limited telemetry. This will help understanding CMEs better by studying their characteristics with improved spatial and temporal resolution.     

First Imaging of Coronal Mass Ejections in the Heliosphere Viewed from Outside the Sun – Earth Line

We show for the first time images of solar coronal mass ejections (CMEs) viewed using the Heliospheric Imager (HI) instrument aboard the NASA STEREO spacecraft. The HI instruments are wide-angle imaging systems designed to detect CMEs in the heliosphere, in particular, for the first time, observing the propagation of such events along the Sun – Earth line, that is, those directed towards Earth. At the time of writing the STEREO spacecraft are still close to the Earth and the full advantage of the HI dual-imaging has yet to be realised. However, even these early results show that despite severe technical challenges in their design and implementation, the HI instruments can successfully detect CMEs in the heliosphere, and this is an extremely important milestone for CME research. For the principal event being analysed here we demonstrate an ability to track a CME from the corona to over 40 degrees. The time – altitude history shows a constant speed of ascent over at least the first 50 solar radii and some evidence for deceleration at distances of over 20 degrees. Comparisons of associated coronagraph data and the HI images show that the basic structure of the CME remains clearly intact as it propagates from the corona into the heliosphere. Extracting the CME signal requires a consideration of the F-coronal intensity distribution, which can be identified from the HI data. Thus we present the preliminary results on this measured F-coronal intensity and compare these to the modelled F-corona of Koutchmy and Lamy (IAU Colloq. 85, 63, 1985). This analysis demonstrates that CME material some two orders of magnitude weaker than the F-corona can be detected; a specific example at 40 solar radii revealed CME intensities as low as 1.7×10⁻¹⁴ of the solar brightness. These observations herald a new era in CME research as we extend our capability for tracking, in particular, Earth-directed CMEs into the heliosphere.     

Automatic Determination of the Conic Coronal Mass Ejection Model Parameters

Characterization of the three-dimensional structure of solar transients using incomplete plane of sky data is a difficult problem whose solutions have potential for societal benefit in terms of space weather applications. In this paper transients are characterized in three dimensions by means of conic coronal mass ejection (CME) approximation. A novel method for the automatic determination of cone model parameters from observed halo CMEs is introduced. The method uses both standard image processing techniques to extract the CME mass from white-light coronagraph images and a novel inversion routine providing the final cone parameters. A bootstrap technique is used to provide model parameter distributions. When combined with heliospheric modeling, the cone model parameter distributions will provide direct means for ensemble predictions of transient propagation in the heliosphere. An initial validation of the automatic method is carried by comparison to manually determined cone model parameters. It is shown using 14 halo CME events that there is reasonable agreement, especially between the heliocentric locations of the cones derived with the two methods. It is argued that both the heliocentric locations and the opening half-angles of the automatically determined cones may be more realistic than those obtained from the manual analysis.     

The ARTEMIS Catalog of LASCO Coronal Mass Ejections

The LASCO-C2 coronagraph aboard the SOHO solar observatory has been providing a continuous flow of coronal images since 1996. Synoptic maps for each Carrington rotation have been built from these images, and offer a global view of the temporal evolution of the solar corona, particularly the occurrence of transient events. Coronal Mass Ejections (CMEs) present distinct signatures thus offering a novel approach to the problem of their identification and characterization. We present in this article an automated method of detection based on their morphological appearance on synoptic maps. It is based on adaptive filtering and segmentation, followed by merging with high-level knowledge. The program builds a catalog which lists the CMEs detected for each Carrington Rotation, together with their main estimated parameters: time of appearance, position angle, angular extent, average velocity and intensity. Our final catalog LASCO-ARTEMIS (Automatic Recognition of Transient Events and Marseille Inventory from Synoptic maps) is compared with existing catalogs, CDAW, CACTUS and SEEDS. We find that, likewise the automated CACTUS and SEEDS catalogs, we detect many more events than the CDAW catalog which is based on visual detection. The total number of detected CMEs strongly depends upon the sensitivity to small, faint and numerous events.     

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.     

基于Faster R-CNN的日冕物质抛射检测方法

日冕物质抛射(Coronal Mass Ejection,CME)是一种强烈的太阳爆发现象,对空间天气和人类生活有巨大的影响,因此,日冕物质抛射检测对预报日冕物质抛射、保障人类的生产生活安全具有重要意义。现有的日冕物质抛射检测多采用人为定义特征和界定阈值等方法。由于人为定义特征不能准确表征日冕物质抛射且具有普适性的阈值难于选择,现有的方法对日冕物质抛射的检测效果有待提高。提出一种基于Faster R-CNN(Faster Region-based Convolutional Neural Networks)的日冕物质抛射检测算法。该方法首先结合CDAW(Coordinated Data Analysis Workshop Data Center),SEEDS(Solar Eruptive Even Detection System)和CACTus(Computer Aoded CME Tracking software package)3个著名的日冕物质抛射目录信息,人工标注了包含9113幅日冕图像的数据集,然后根据日冕物质抛射的图像特征较自然图像少、目标尺寸与自然图像有差异等特点,在特征提取和锚点选择方面对Faster R-CNN进行改进。以2007年6月的日冕物质抛射标注数据为测试集,本文算法检出了全部22个强日冕物质抛射事件和151个弱日冕物质抛射事件中的138个,对日冕物质抛射事件的中心角和角宽度等特征参数的检测误差分别在5°和10°以内。     

Geometric Localization of CMEs in 3D Space Using STEREO Beacon Data: First Results

The geometric localization technique (Pizzo and Biesecker, Geophys. Res. Lett. 31, 21802, 2004) can readily be used with Solar Terrestrial Relations Observatory (STEREO) Space Weather Beacon data to observe coronal mass ejection (CME) propagation within three-dimensional space in near-real time. This technique is based upon simple triangulation concepts and utilizes a series of lines of sight from two space-based observatories to determine gross characteristics of CMEs, such as location and velocity. Since this work is aimed at space weather applications, the emphasis is on use of COR2 coronagraph data, which has a field of view from 2.5R _⊙ to 15R _⊙; this spatial coverage allows us to observe the early temporal development of a CME, and hence to calculate its velocity, even for very fast CMEs. We apply this technique to highly-compressed COR2 beacon images for several CMEs at various spacecraft separation angles: 21 August 2007, when the separation angle between the two spacecraft was 26°; 31 December 2007 and 2 January 2008, when the separation angle was 44°; and 17 October 2008, when the spacecraft separation was 79°. We present results on the speed and direction of propagation for these events and discuss the error associated with this technique. We also compare our results to the two-dimensional plane-of-sky speeds calculated from STEREO and SOHO.     

A Comparative Study of Coronal Mass Ejections with and Without Magnetic Cloud Structure near the Earth: Are All Interplanetary CMEs Flux Ropes?

An outstanding question concerning interplanetary coronal mass ejections (ICMEs) is whether all ICMEs have a magnetic flux rope structure. We test this question by studying two different ICMEs, one having a magnetic cloud (MC) showing smooth rotation of magnetic field lines and the other not. The two ICMEs are chosen in such a way that their progenitor CMEs are very similar in remote sensing observations. Both CMEs originated from close to the central meridian directly facing the Earth. Both CMEs were associated with a long-lasting post-eruption loop arcade and appeared as an elliptical halo in coronagraph images, indicating a flux rope origin. We conclude that the difference in the in-situ observation is caused by the geometric selection effect, contributed by the deflection of flux ropes in the inner corona and interplanetary space. The first event had its nose pass through the observing spacecraft; thus, the intrinsic flux rope structure of the CME appeared as a magnetic cloud. On the other hand, the second event had the flank of the flux rope intercept the spacecraft, and it thus did not appear as a magnetic cloud. We further argue that a conspicuous long period of weak magnetic field, low plasma temperature, and density in the second event should correspond to the extended leg portion of the embedded magnetic flux rope, thus validating the scenario of the flank-passing. These observations support the idea that all CMEs arriving at the Earth include flux rope drivers.     

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.     

Propagation Characteristics of CMEs Associated with Magnetic Clouds and Ejecta

We have investigated the characteristics of magnetic cloud (MC) and ejecta (EJ) associated coronal mass ejections (CMEs) based on the assumption that all CMEs have a flux rope structure. For this, we used 54 CMEs and their interplanetary counterparts (interplanetary CMEs: ICMEs) that constitute the list of events used by the NASA/LWS Coordinated Data Analysis Workshop (CDAW) on CME flux ropes. We considered the location, angular width, and speed as well as the direction parameter, D. The direction parameter quantifies the degree of asymmetry of the CME shape in coronagraph images, and shows how closely the CME propagation is directed to Earth. For the 54 CDAW events, we found the following properties of the CMEs: i) the average value of D for the 23 MCs (0.62) is larger than that for the 31 EJs (0.49), which indicates that the MC-associated CMEs propagate more directly toward the Earth than the EJ-associated CMEs; ii) comparison between the direction parameter and the source location shows that the majority of the MC-associated CMEs are ejected along the radial direction, while many of the EJ-associated CMEs are ejected non-radially; iii) the mean speed of MC-associated CMEs (946 km?s^?1) is faster than that of EJ-associated CMEs (771 km?s^?1). For seven very fast CMEs (≥?1500 km?s^?1), all CMEs with large D (≥?0.4) are associated with MCs and the CMEs with small D are associated with EJs. From the statistical analysis of CME parameters, we found the superiority of the direction parameter. Based on these results, we suggest that the CME trajectory essentially determines the observed ICME structure.     

Entering The Era Of Automated Cme Recognition: A Review Of Existing Tools

We consider the problem of the observational identification of CMEs. The ever growing importance of space weather has led to new requirements on the timeliness and objectiveness of CME detection. It is not sufficient any more to simply detect CMEs, a complete set of characteristics (speed, direction, mass, chirality) must be reported as soon as possible to estimate its geoeffectiveness. Recent developments in (solar) feature recognition greatly improved the ability to address these new needs. Progress was achieved in automating the detection of CMEs in coronagraphic data. This has led to near-real-time messages alerting the space weather community day and night. In attempting to generate ever-prompter alerts, we can employ a far broader set of solar observations than coronagraphic data alone. At present an extensive set of automatic recognition tools exists for a number of CME-related phenomena occurring in the lower corona. This paper deals with detection techniques for disappearing filaments in Hα images, dimmings, EIT waves and erupting prominences in radio data. We believe that incorporating all automatically generated alerts into one report per CME can provide valuable CME information, especially when no coronagraphic images are available. This paper is thus a quest to reach a maximal success rate with the help of an integrated system of tools acting on a variety of data. Future grid-technology systems will greatly facilitate this.     

Deflection and Rotation of CMEs from Active Region 11158

Between 13 and 16 February 2011, a series of coronal mass ejections (CMEs) erupted from multiple polarity inversion lines within active region 11158. For seven of these CMEs we employ the graduated cylindrical shell (GCS) flux rope model to determine the CME trajectory using both Solar Terrestrial Relations Observatory (STEREO) extreme ultraviolet (EUV) and coronagraph images. We then use the model called Forecasting a CME’s Altered Trajectory (ForeCAT) for nonradial CME dynamics driven by magnetic forces to simulate the deflection and rotation of the seven CMEs. We find good agreement between ForeCAT results and reconstructed CME positions and orientations. The CME deflections range in magnitude between \(10^{\circ }\) and \(30^{\circ}\). All CMEs are deflected to the north, but we find variations in the direction of the longitudinal deflection. The rotations range between \(5^{\circ}\) and \(50^{\circ}\) with both clockwise and counterclockwise rotations. Three of the CMEs begin with initial positions within \(2^{\circ}\) from one another. These three CMEs are all deflected primarily northward, with some minor eastward deflection, and rotate counterclockwise. Their final positions and orientations, however, differ by \(20^{\circ}\) and \(30^{\circ}\), respectively. This variation in deflection and rotation results from differences in the CME expansion and radial propagation close to the Sun, as well as from the CME mass. Ultimately, only one of these seven CMEs yielded discernible in situ signatures near Earth, although the active region faced toward Earth throughout the eruptions. We suggest that the differences in the deflection and rotation of the CMEs can explain whether each CME impacted or missed Earth.