The SEEDS team at George Mason University is:
Jie Zhang (team leader), Oscar Olmedo (leading developer), Harry Wechsler, Art Poland, Kirk Borne

The Solar Eruptive Event Detection System (SEEDS) is a software package under development to automatically detect, characterize and classify transient/eruptive solar events, including CMEs, coronal dimmings, and flares, using image processing, machine learning and data mining technique.

As the first stage of development an algorithm was implemented that detects and tracks CMEs, as seen in LASCO C2 images and soon with STEREO COR2 images.

Refer to the publication for details
Olmedo, O., Zhang, J., Wechsler, H., Poland, A., Borne, K.: 2008, Solar Phys. 248, 485.
"Automatic Detection and Tracking of Coronal Mass Ejections in Coronagraph Time Series"

The advantage of this algorithm is that detections are made directly on running-difference images and many parameters are generated that can be used to classify CME events. The most important of these patameters are the velocity, acceleration, and position angle, these are seen in the columns of the monthly detection tables in the SEEDS monthly catalog. Comparing the SEEDS CME detections to the monthly catalogs prepared by a visual manual method (the CDAW catalog), approximatly 75% of the CMEs are recovered. As is the case with other automated detection methods (CACTUS), an excessive number of events are seen, where most of these are related to, either pre or post, outflows of CMEs and a few are real events, but were unreported in the visual catalogs.
There are three main stages in the algorithm, pre-processing, detection, and tracking. The pre-processing stage begins with transforming the Coronograph image into polar coordinates where the x axis represents angle and the y axis the height. A running difference is made of the data and each running difference image is normalized such that its mean is zero. The next step is the detection, first a projection of the transformed image is taken in the x-axis (Fig. 1), with thresholds, outstanding angles are found (Fig. 2). Region Growing and Thresholding segentation determine the angular width of the cme as well as an approximate outline of the CME (Fig. 3). From here the CME can be tracked in subsequent images.

Fig. 1

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 8

Fig. 9

Fig. 6

Fig. 7

Fig. 10

Fig. 11

Curators: Jie Zhang (jzhang7 AT; Suman Dhakal (sdhakal2 AT;
Last updated: 2020/04