HR: 1330h
AN: SP23A-03    [Abstracts]
TI: Automatic Detection of Corona Mass Ejections Using Pattern Recognition Techniques
AU: * Qu, M
EM: qm3@njit.edu
AF: NJIT, 323 Martin Luther King Boulevard, 403 Tiernan Hall, newark, nj 07102 United States
AU: Shih, F Y
EM: shih@njit.edu
AF: NJIT, 323 Martin Luther King Boulevard, 403 Tiernan Hall, newark, nj 07102 United States
AU: Jing, J
EM: jj4@njit.edu
AF: NJIT, 323 Martin Luther King Boulevard, 403 Tiernan Hall, newark, nj 07102 United States
AU: Wang, H
EM: haimin@flare.njit.edu
AF: NJIT, 323 Martin Luther King Boulevard, 403 Tiernan Hall, newark, nj 07102 United States
AB: We present an automatic method to detect and categorize Corona Mass Ejections (CMEs) using the advanced pattern recognition technique. Our method is applied to LASCO (Large Angel Spectrometric Coronagraph) C2 and C3 images. CMEs are very complex features that are varied from time to time. The noises on the images also reduce the accuracy of the CME detections. Therefore, we propose different procedures for different structural CMEs that are described by Howard et al. (1985). There are three steps in our detection and characterization. (1)Preprocess: our program correlates C2 with C3 images, normalizes the images to the same contrast level and remove noises. Three original images are used to produce two difference images, and images are reformatted to angular images. (2) Characterization: automatic thresholding and morphology methods are used to segment and identify the principal objects from the images. CMEs are characterized according to their intensity, height, span, velocity and mass. (3) Classification: by detail study, twenty features are proposed to represent a CME. With these twenty features, the SVM classifiers are able to detect CMEs and categorize CMEs.
DE: 7599 General or miscellaneous
SC: Solar Physics Division - AAS [SP]
MN: 2005 Joint Assembly