Research On Coronal Mass Ejection Detection Method Based On Spatiotemporal Saliency

Coronal Mass Ejections(CME)is a multi-scale energy explosion that frequently occurs on the surface of the Sun within a few minutes to several hours.CME spreads in interplanetary space to form Interplanetary Coronal Mass Ejection(ICME).When ICME reaches the earth's space environment,it causes earth satellite operation failures and ground power transmission collapse,which seriously endangers the normal production and life of human beings.The detection of CME can predict the time when ICME arrives on the earth and help humans reduce the impact of CME.The paper introduces the research background and saliency of CME detection,briefly summarizes the current situation of CME detection research work at home and abroad,and points out the problems existing in the current CME detection research work.Then briefly introduce the background knowledge of CME and the authoritative CME detection catalog,including CME characteristic physical parameters,manual detection methods and corresponding catalog: Coordinated Data Analysis Workshop(CDAW),automatic detection method: Computer Aided CME Tracking Software package(CACTus),Solar Eruptive Event Detection System(SEEDS)and corresponding catalog.In view of the fact that the existing CME detection methods can not achieve the accurate detection of CME well,ignore the CME movement time information,there are problems such as error detection and missed detection,and the application research is carried out around CME detection.The paper conducts theoretical analysis and research on the visual saliency detection algorithm,makes full use of its performance advantages,uses the spatial saliency and motion time information of the coronal mass ejection in the coronal image sequence,and proposes a CME detection based on spacetime saliency method.Use the spatial saliency detection algorithm based on Saliency Filters(SF)to obtain the CME spatial saliency map,and then use the improved threeframe difference method to combine the space-time saliency to detect the CME based on the movement time dimension information of the CME.In order to test the effectiveness of the method,two corona image sequences that occurred at different times were selected for verification,and the manual catalog CDAW was used as a benchmark reference,and the number of CME detections,accuracy,and characteristic parameters were compared with the manual catalog CDAW,automatic catalog CACTus,and SEEDS.The experimental results show that when the manual catalog CDAW is used as the benchmark reference,the number of CMEs successfully detected by the proposed method is 23%and 18%higher than that of CACTus and SEEDS respectively in time period I,and higher in time period II.32%,40%.In addition,it is closer to the CDAW catalog reference value than CACTus and SEEDS in the detection of CME characteristic parameters.