Abstract
Filaments are omnipresent features in the solar chromosphere. Their location, properties and time evolution can provide important information about changes in solar activity and assist the operational space weather forecast. Therefore, filaments have to be identified in full-disk images and their properties extracted from these images. Manual extraction is tedious and takes too much time; extraction with morphological image processing tools produces a large number of false-positive detections. Automatic object detection, segmentation, and extraction in a reliable manner allows us to process more data in a shorter time. The Chromospheric Telescope (ChroTel), Tenerife, Spain, the Global Oscillation Network Group (GONG), and the Kanzelhöhe Solar Observatory (KSO), Austria, provide regular full-disk observations of the Sun in the core of the chromospheric Hα absorption line. We present a deep learning method that provides reliable extractions of solar filaments from Hα filtergrams. First, we train the object detection algorithm YOLOv5 with labeled filament data of ChroTel Hα filtergrams. We use the trained model to obtain bounding-boxes from the full GONG archive. In a second step, we apply a semi-supervised training approach, where we use the bounding boxes of filaments, to learn a pixel-wise classification of solar filaments with u-net. Here, we make use of the increased data set size which avoids overfitting of spurious artifacts from the generated training masks. With the resulting filament segmentations, physical parameters such as the area or tilt angle can be easily determined and studied. This we demonstrate in one example, where we determine the rush-to-the pole for Solar Cycle 24 from the segmented GONG images. In a last step, we apply the filament detection to Hα observations from KSO which demonstrates the general applicability of our method to Hα filtergrams.
Recorded video
https://science-media.org/video/354