基于监督学习的日冕暗化检测与提取算法

Coronal Dimming Detecting and Extracting Algorithm Based on Supervised Learning

日冕物质抛射(CME)是空间灾害天气的重要驱动源,而日冕暗化(dimming)被认为是CME初发的主要表征,对理解和预测CME具有重要作用。基于极紫外成像望远镜(EIT)和大气成像仪(AIA)的观测数据,实现了图像中日冕暗化现象的检测与提取。通过分析差分图中与暗化现象相关的图像统计特征,采用Adaboost分类算法检测暗化现象的发生,进而分割出日冕暗化区域。实验表明,提出的算法较现有算法能更准确有效地检测和提取日冕暗化区域,为分析日冕暗化特性提供了研究基础。

Coronal mass ejections （CMEs）, which release huge quantities of matter and electromagnetic radiation into space above the sun＇s surface, are considered as one of the driven sources of space weather. Coronal dimming is now viewed as the important characteristic of CME. Dimming can help understand, predict and locate the occurrence of CME. Based on the observed data from extreme ultra-violet imaging telescope （EIT） and atmospheric imaging assembly （AIA）, this paper implemented the coronal dimming detection and extraction, By analyzing the statistical characteristics of the difference images related to dimming, we applied Adaboost classification algorithm into dimming detection, and then segmented the coronal dimming region. The experiment results show that the proposed algorithm can effectively detect and extract the coronal dimming areas. Our work establishes the basis for analysis of coronal dimming features.