Generating Space-based SDO/HMI-like Solar Magnetograms from Ground-based Hα Images by Deep Learning

Abstract Recently, the method of estimating magnetic field through monochromatic images by deep learning has been proposed, demonstrating good morphological similarity but somewhat poor polarity consistency relative to real observation. In this paper, we propose estimate from H α using a conditional generative adversarial network (cGAN) as basic framework. The Global Oscillation Network Group are used inputs and line-of-sight magnetograms Helioseismic Magnetic Imager (HMI) targets. First, train cGAN model (Model A) with shuffling training data. However, estimated polarities not very consistent observations. Second, improve accuracy polarities, B) chronological HMI images, which can implicitly exploit constraint time-series observation generate more accurate polarities. We compare generated target evaluate two models. It be observed that Model B better than A. To quantitatively measure consistency, new metric called pixel-to-pixel (PPA). With respect PPA, is superior This work gives us an insight exploited organizing data chronologically, conclusion also applied other similar tasks related