基于期望最大算法的空间事件及异常值探测

美国ELSET数据库提供的TLE数据是目前使用最广泛的数据，在热层大气密度反演、弹道系数估计、碰撞预警等领域具有重要作用。受空间环境扰动、空间事件以及TLE产生过程等共同影响，ELSET数据库包含大量亟待清理的异常值和识别的空间事件，例如发布错误的TLE、轨道根数异常和Bstar异常。现有方法在清理异常轨道根数时缺乏统一性，需要使用不同的技术，清理流程较为繁杂，并且仅适用于特定轨道区域的少数目标。为克服现有方法的弊端，提出了一种基于期望最大算法的滑动窗口–多项式拟合预报方法，对含有轨道机动的碎片以及受空间环境影响的碎片进行异常值与空间事件探测。研究表明，该方法能够灵活处理不同空间环境下的异常值与空间事件探测，具有普适性，适用于所有轨道碎片。 

Space Event and Outlier Detection Based on Expectation Maximization Algorithm

The United States provide Element Sets (ELSET) database in Two-Line Element (TLE) format for public use, which plays an important role in the inversion of atmospheric density in the thermosphere, ballistic coefficient estimation, early-warning and so on. Due to large uncertainties existing in the TLE generation process, space environment changes and space events, ELSET database contains a large number of abnormal TLE data to be filtered, such as corrected TLE, orbital element outlier and Bstar outlier. The existing methods to filter out the outliers lack general applicability and are very complicated, which are only applicable to a few space targets in certain orbit regions. To overcome the shortcomings of the existing methods, a filtering method is proposed based on Expectation Maximization (EM) algorithm employing a sliding window and polynomial fitting method, which can detect outliers for different orbital elements and space events. The research shows that the algorithm can effectively single out the outliers in TLE sequences and is suitable for all orbital debris.