Automated Detection and Identification of Solar Filaments and Sunspots

We developed a procedure for the automatic detection and identification of filaments and their disappearance. Full-disk Hα images from the Big Bear Solar Observatory (BBSO) in California are used as the data set for our procedure. Solar images are randomly selected starting from January 1, 1999 to September 1, 2004. We present an automatic solar filament detection procedure using advanced image enhancement, segmentation, pattern recognition and mathematical morphology. This procedure not only provides the detection results of filaments, but also identifies the spines, footpoints and disappearances of filaments. Low contrast filaments are emphasize and sharpen by the stabilized inverse diffusion equation (SIDE) which was introduced by Pollak et al. (2000). Adaptive image segmentation techniques are used for selecting the threshold based on the edge and local information. To distinguish sunspot from filaments, an efficient feature-based classifier, the Support Vector Machine (SVM), is utilized. Detail filament identification is achieved by morphological thinning, pruning and adaptive edge linking methods. Finally, the filament disappearances are detected by comparing the spine and footpoints of the filaments on two consecutive days. Comparing to Gao et al. (2002) and Shih and Kowalski (2003), our procedure utilizes the image enhancement techniques to enhance the low contrast filaments, and apply advanced pattern recognition and morphology techniques to identify filament and sunspots. Our work has shown the better and more complete results than other work on the automatic filament detection.