This paper presents a generic method to automatically segment and characterize solar filaments from various Hα full-disk solar images, obtained by different solar observatories, with different dynamic ranges and statistical properties. First, a cascading Hough circle detector is designed to find the center location and radius of the solar disks. Second, polynomial surface fitting is adopted to correct unbalanced luminance. Third, an adaptive thresholding method is put forward to segment solar filaments. Finally, for each piece of a solar filament, its centroid location, area, and length are characterized, in which morphological thinning and graph theory are used for identifying the main skeletons of filaments. To test the performance of the proposed methods, a dataset composed of 125 Hα images is considered. These images were obtained by four solar observatories from January 2000 to May 2010, one image per month. Experimental results show that the accuracy rate is above 95% as measured by filament number and above 99% as measured by filament area, indicating that only a few tiny filaments are not detected.
All Science Journal Classification (ASJC) codes
- Astronomy and Astrophysics
- Space and Planetary Science