Abstract: Arc aurora is an important form of dayside aurora. Its distribution and occurrence reveals magnetospheric pulsation of geomagnetic activity. It can help people to study the physical occurring mechanisms of arc aurora by the research on the occurrence time of arc aurora. According to the characteristics of arc aurora sequence, this paper proposes a new feature extraction method based on RLDP (ring local directional pattern), and gets the starting and ending time of aurora sequence by using the STSC (self-tuning spectral clustering). Compared with the image sequences labeled by manual, the experimental results show that the proposed method has better performance both on recall rate and precision rate.

Key words: arc aurora sequence, ring local directional pattern (RLDP), self-tuning spectral clustering (STSC)

摘要： 弧状极光作为日侧极光的一个重要形态，它的分布和发生揭示了地磁活动和磁层中的脉动现象。对弧状极光发生起始时刻的判定能有效帮助人们研究弧状极光发生的物理机制。针对弧状极光序列的特征，提出了基于环形局部方向模式（ring local directional pattern，RLDP）的特征表示新方法，再利用自动调节谱聚类（self-tuning spectral clustering，STSC）算法得到极光序列的发生和结束时刻。将检测结果和人工标记的图像序列进行对比，基于环形局部方向模式的特征表示方法得到了较高的查全率和查准率。

关键词: 弧状极光序列, 环形局部方向模式（RLDP）, 自动调节谱聚类（STSC）