We modeled the detection of Coronal Mass Ejections(CMEs) as the classification of the brightest block in the current running difference image.Because CME has multiple features and amorphous,We design ensemble classification schemes to fuse the classification results from multiple features that include Gray-level Histogram,Haralick texture and HOG features.And We design our detection algorithm by AdaBoost and decision trees.
The following is our detection results.Data we used are taken from May 1 to May 31,2007. Our catalog illustrates the results from two aspects that include start point of the CME event and continuous frames.