Magnetic Classification Of Sunspot Group Wilson Mountain Based On CornerNet-Saccade

Closely related to solar activities,such as flares and coronal mass ejections,sunspots are a typical manifestation of the strong magnetic field on the surface of the sun.These activities not only disrupt the earth's atmosphere and affect the ground short-wave radio communication but also produce "magnetic storms" and other hazards.According to studies,the more complex morphological structures of sunspot groups and magnetic polarity within the corresponding active regions are,the higher probability of flares in the regions is.In the Mount Wilson sunspot magnetic classification scheme,the ??? is the most closely associated with flares,above which most M-class and X-class flares occur.The relation between sunspot groups and solar flares is of great significance in the study of the Mount Wilson sunspot magnetic classification algorithm.As good results are obtained in the field of object detection by applying the deep learning model CornerNet-Saccade,the author studies the Mount Wilson sunspot magnetic classification scheme on the basis of it.This paper establishes a reliable data set of the Mount Wilson sunspot magnetic classification,preprocesses about 2,400 full heliographic charts of the sun,and uses Label Me labeling tool to classify and label the sunspot groups.A total of more than10,000 samples are labeled.It has been repeatedly confirmed with relevant experts from Yunnan Observatory.Aiming at the morphological characteristics of the sunspot group and the magnetic field polarity of the corresponding active area,a CNSS(CornerNet-Saccade-Sunspot)model based on CornerNet-Saccade was proposed and successfully applied to the Wilson Mountain magnetic classification task of the sunspot groups.The detection results show that the precision,the recall,and the m AP are 95%,and 93% respectively and the detection precision has increased by 6%.The innovation of the paper is the establishment of a reliable data set of Mount Wilson sunspot magnetic classification,which is used to train and test models when the Mount Wilson sunspot magnetic classification is carried out.Besides,in terms of the morphological characteristics of sunspot groups and the magnetic polarity within the corresponding active regions,the CNSS model based on CornerNet-Saccade is proposed in the paper greatly improves the detection performance.