No abstract

The Atmospheric Imaging Assembly (AIA) onboard the Solar Dynamics Observatory (SDO) provides full-disk solar images with high temporal cadence and spatial resolution over seven extreme ultraviolet (EUV) wavebands. However, as violent solar flares happen, images captured in EUV wavebands may have saturation in active regions, resulting in information loss. In this paper, we propose a deep learning model to restore the lost signal in saturated regions by referring to both unsaturated/normal regions within a solar image and statistical probability model of massive normal solar images. The proposed model, namely mixed convolution network (MCNet), is established over conditional generative adversarial network (GAN) and the combination of partial convolution (PC) and validness migratable convolution (VMC). These two convolutions were originally proposed for image inpainting. In addition, they are implemented only on unsaturated/valid pixels, followed by certain compensation to compensate the deviation of PC/VMC relative to normal convolution. Experimental results demonstrate that the proposed MCNet achieves favorable desaturated results for solar images and outperforms the state-of-the-art methods both quantitatively and qualitatively.

A solar radio spectrometer records solar radio radiation in the radio waveband. Such solar radio radiation spanning multiple frequency channels and over a short time period could provide a solar radio spectrum which is a two dimensional image. The vertical axis of a spectrum represents frequency channel and the horizontal axis signifies time. Intrinsically, time dependence exists between neighboring columns of a spectrum since solar radio radiation varies continuously over time. Thus, a spectrum can be treated as a time series consisting of all columns of a spectrum, while treating it as a general image would lose its time series property. A recurrent neural network (RNN) is designed for time series analysis. It can explore the correlation and interaction between neighboring inputs of a time series by augmenting a loop in a network. This papermakes the first attempt to utilize an RNN, specifically long short-termmemory (LSTM), for solar radio spectrum classification. LSTM can mine well the context of a time series to acquire more information beyond a non-time series model. As such, as demonstrated by our experimental results, LSTM can learn a better representation of a spectrum, and thus contribute better classification.