Computer Vision for the Solar Dynamics Observatory (SDO)

Abstract

In Fall 2008 NASA selected a large international consortium to produce a comprehensive automated feature-recognition system for the Solar Dynamics Observatory (SDO). The SDO data that we consider are all of the Atmospheric Imaging Assembly (AIA) images plus surface magnetic-field images from the Helioseismic and Magnetic Imager (HMI). We produce robust, very efficient, professionally coded software modules that can keep up with the SDO data stream and detect, trace, and analyze numerous phenomena, inThe Solar Dynamics Observatory Guest Editors: W. Dean Pesnell, Phillip C. Chamberlin, and Barbara J. Thompson. cluding flares, sigmoids, filaments, coronal dimmings, polarity inversion lines, sunspots, Xray bright points, active regions, coronal holes, EIT waves, coronal mass ejections (CMEs), coronal oscillations, and jets. We also track the emergence and evolution of magnetic elements down to the smallest detectable features and will provide at least four full-disk, nonlinear, force-free magnetic field extrapolations per day. The detection of CMEs and filaments is accomplished with Solar and Heliospheric Observatory (SOHO)/Large Angle and Spectrometric Coronagraph (LASCO) and ground-based Hα data, respectively. A completely new software element is a trainable feature-detection module based on a generalized imageclassification algorithm. Such a trainable module can be used to find features that have not yet been discovered (as, for example, sigmoids were in the pre-Yohkoh era). Our codes will produce entries in the Heliophysics Events Knowledgebase (HEK) as well as produce complete catalogs for results that are too numerous for inclusion in the HEK, such as the X-ray bright-point metadata. This will permit users to locate data on individual events as well as carry out statistical studies on large numbers of events, using the interface provided by the Virtual Solar Observatory. The operations concept for our computer vision system is that the data will be analyzed in near real time as soon as they arrive at the SDO Joint Science Operations Center and have undergone basic processing. This will allow the system to produce timely space-weather alerts and to guide the selection and production of quicklook images and movies, in addition to its prime mission of enabling solar science. We briefly describe the complex and unique data-processing pipeline, consisting of the hardware and control software required to handle the SDO data stream and accommodate the computer-vision modules, which has been set up at the Lockheed-Martin Space Astrophysics Laboratory (LMSAL), with an identical copy at the Smithsonian Astrophysical Observatory (SAO).

Application of Support Vector Machine to the Prediction of Geo-Effective Halo Cmes

“…By using CME speed and X-ray flare class as input features of the SVM, we have obtained statistical parameters for the best model: Accuracy=0.66,PODy=0.76,PODn=0.49,FAR=0.72,Bias=1.06,CSI=0.59,and TSS=0.25. In recent times, the needs of machine learning are growing because it can deal with large amount of data from space missions. For example, automated systems using machine learning algorithms have been developed for the solar dynamics observatory (SDO) mission (Martens et al 2009;Attrill et al 2010). From the present study together with the previous ones, the most important advantage of the SVM for space weather forecast is that we can select the best model among many possible models by changing various kernels and their parameters.…”

“…Therefore machine learning technology has been employed for space weather applications in the following two aspects: space weather prediction (Al-Omari et al 2010;Chen et al 2010;Colak et al 2009;Gavrishchaka et al 2001;He et al 2008;Li et al 2007;Liu Corresponding Author : Y.-J. Moon et al 2011;Olmedo et al 2005;Qahwaji et al 2007Qahwaji et al , 2008Yuan et al 2011) and solar feature identification (Henwood et al 2010;Labrosse et al 2010;Quaalude et al 2003Quaalude et al , 2005Martens et al 2009).…”

Hemispheric Patterns in Filament Chirality and Sigmoid Shape over the Solar Cycle

“…The usual attribution criterion is that two more barbs have to be of one bearing versus the opposite one to assign filament chirality. Georgoulis' "Sigmoid Sniffer" (see Martens et al 2012) likewise detects whether the contour of a sigmoid has an "S" (forward) or a "Z"-shape (reverse).…”

“…One of these teams is headed by the first author of this paper and has produced 16 modules for the detection and analysis of different solar features (Martens et al 2012), see also the more up-to-date website †. The analysis of the metadata from two of these modules, the "Advanced Automated Filament Detection and Characterization Code (AAFDCC)", see Bernasconi, Rust & Hakim (2005), and the "Sigmoid Sniffer" (Martens et al 2012) will be presented in this paper, and the results contain surprises.…”

“…The analysis of the metadata from two of these modules, the "Advanced Automated Filament Detection and Characterization Code (AAFDCC)", see Bernasconi, Rust & Hakim (2005), and the "Sigmoid Sniffer" (Martens et al 2012) will be presented in this paper, and the results contain surprises.…”