Proper elements represent a dynamical fingerprint of an object’s inherent state and have been used by small-body taxonomists in characterizing asteroid families. Being linked to the underlying dynamical structure of orbits, Celletti, Pucacco, and Vartolomei have recently adopted these innate orbital parameters for the association of debris from breakup or collision into its parent satellite. Building from this rich astronomical heritage and recent foundations, we introduce an unsupervised learning method—density-based spatial clustering of applications with noise (DBSCAN)—to determine clusters of orbital debris in the space of proper elements. Data is taken from the space-object catalog of trackable Earth-orbiting objects in the form of two-line element sets. Proper elements for debris fragments in low-Earth orbit are computed using an ad hoc numerical scheme, akin to the state-of-the-art Fourier-series-based synthetic method for the asteroid domain. Given the heuristic nature of classical DBSCAN, we investigate the use of neural networks, trained on known families, to augment DBSCAN into a classification problem and apply it to analyst objects of unknown origin.

固有元素代表了物体固有状态的动态指纹，并已被小天体分类学家用来描述小行星家族的特征。与轨道的潜在动力学结构相关联，切莱蒂、普卡科和瓦尔托洛梅最近采用了这些固有轨道参数，将破碎或碰撞产生的碎片与其母卫星关联起来。基于丰富的天文遗产和最新的基础，我们引入了一种无监督学习方法——基于密度的噪声应用空间聚类（DBSCAN）——来确定适当元素空间中的轨道碎片集群。数据取自可跟踪地球轨道物体的空间物体目录，采用两行元素集的形式。低地球轨道碎片碎片的适当元素是使用专门的数值方案计算的，类似于小行星域最先进的基于傅立叶级数的合成方法。考虑到经典 DBSCAN 的启发式性质，我们研究了使用在已知家族上训练的神经网络，将 DBSCAN 增强为分类问题，并将其应用于分析未知来源的对象。