Using Deep Learning to Detect and Trace H-alpha Fibrils

We present a new deep learning method, named FibrilNet, for detecting and tracing chromospheric fibrils in Hα images of solar observations. Our method consists of a data pre-processing component that prepares training data from a physics-based tool, a deep learning model implemented as a Bayesian convolutional neural network for probabilistic image segmentation with uncertainty quantification, and a post-processing component with a polynomial regression model for preparing tracing results of both fibrils and their orientation. The FibrilNet tool is applied to data from the 1.6 m Goode Solar Telescope equipped with high-order adaptive optics at the Big Bear Solar Observatory with promising results.