Solar Activity Recognition Based On The Knowledge Base Of The Japanese Physics Event

The development of content-based image retrieval(CBIR)has been nearly thirty years,which has been widely used in astronomical image processing.With the development of multiband solar observation technologies,people can get higher and higher resolution of the sun's image,which greatly prompt the development of solar activity research.As the amount of astronomical image is raising rapidly,also bring great challenges to the data storage of astronomical image accordingly.How to detect and identify useful solar activities automatically from full-disk image is another difficult problem in the astronomical image processing scope.Based on above two problems,this thesis includes the following two aspects.First,This thesis proposes a target detection method for solar activity based on the structure of square grid(GBTD).In this method,the solar image is divided into a square grid structure with the same size,based on multiply-thresholds selection strategy and GBTD strategy to segment the target area and background area.After segmenting six kinds of solar activities,total of 2172 solar activity areas.The experimental results show that this method get the satisfactory performance in accuracy and time-cost,as well as an obvious anti-jamming performance for image noise.GBTD method provides a general image segmentation method for the investigation of various types of solar image,also to provide a feasible method for solving the problem of mass astronomical data storage.Second,we study the correlation of six kinds of solar image features,and give combinations of features for every solar activity.The method in the paper realizes the precise positioning and efficient recognition,which laid the foundation for further work.Besides,extracting particular combination of features from different solar activities regions,which provide a practical approach of constructing a condensed feature set for CBIR system.Third,Benefiting from real-time observation data provided by Heliophysics Event Knowledgebase(HEK)of Solar Dynamics Observatory(SDO).This paper proposes a recognition method of solar activities based on HEK.In this method,we gather six kinds of solar activities datas(date,location,area),then we design a scaling transform model for full-disk solar image of corresponding date.Combining location data and area data,we design different gradient thresholds to segment different solar activities regions.Two kinds of boundary identification method are used to locate and recognize solar activities regions respectively.The method in the paper realizes the precise positioning and efficient recognition,which laid the foundation for further work.