Research On Automatic Recognition Of Solar Active Regions

Solar active regions are the source of most solar eruption activities.Their characteristic parameters are essential to predict solar activities.With a rapid increase in the size of solar image data archives,the automated detection and verification for solar active regions is of great interest to space weather forecast community.Based on the Space-Weather HMI Active Region Patch(SHARP)data of the Helioseismic and Magnetic Imager(HMI)in the Solar Dynamics Observatory(SDO),we have developed an automatic recognition procedure for solar active regions.The contents of this thesis are as follows:1.The thesis has constructed an automatic sunspot detection procedure by a mathematical morphology tool,which can calculate sunspot group area and sunspot number.By comparing out results to those obtained from Solar Region Summary compiled by NOAA/SWPC,it is found that both the sunspot group area and sunspot number computed with our algorithm agree well with the active region values compiled by SWPC,and the corresponding correlation coefficients of sunspot group area and sunspot number are 0.77 and 0.79,respectively;2.The McIntosh classification of sunspot groups is identified by the Convolution Neural Network Inception V3 model,which is based on the TensorFlow platform.The three components Z,p,and c of the McIntosh type are classified and trained separately by three different image classifiers.It turns out that the recognition accuracy is 75.4%,79.2%,and 82.2% for Z,p,and c component,respectively.