Exploration of U-Net in Automated Solar Coronal Loop Segmentation

This paper presents a deep convolutional neural network (CNN) based method that automatically segments arc- like structures of coronal loops from the intensity images of Sun’s corona. The method explores multiple U-Net architecture variants which enable segmentation of coronal loop structures of active regions from NASA’s Solar Dynamic Observatory (SDO) imagery. The effectiveness of the method is evaluated through experiments on both synthetic and real images, and the results show that the method segments the coronal loop structures accurately.