Detection and classification of sunspots via deep convolutional neural network
Channabasava Chola, J V Biabl Benifa
Abstract
Sunspots are known to be the most prominent feature of the solar photosphere. Solar activities play a vital role in Space weather which greatly affects the Earth's environment. The appearance of sunspots determines the solar activities and being observed from early eighteenth century. In this work, we have implemented a deep learning model which automatically detects sunspots from MDI and HMI image datasets. Proposed model uses Alexnet based deep convolutional networks to generate promising deep hierarchical features and proposed deep learning approach achieved excellent classification accuracies. Also, model has shown the improved result with MDI data set which is equal to 99.71%, 100%, 100%, and 100 for accuracy, precision, recall, and F-score respectively. This is to construct and build robust and reliable event recognition system to monitor solar activities which are crucial to understanding space weather and for physicists it is an aid for their research.