Space object classification based on non-conservative force

The increasingly crowded space environment necessitates enhanced Space Situational Awareness (SSA) capabilities. In the SSA system, the essential task is to classify space objects such as operational satellites and defunct debris for various purposes.Traditional approaches often rely on single observational data like light curve, radar cross section, or optical image. However, orbital parameters, which are updated frequently and cover a much larger population of resident space objects, provide a complementary and information-rich data source. In this study, we explore a novel approach for space object classification based on orbital parameters. We first derive the Non-Conservative Force (NCF) accelerations from the orbital parameters and then extract a set of features from the NCF time series. These features are subsequently used to train several conventional classification algorithm, including support vector machine, k-nearest neighbors, and decision tree. The accuracy of the NCF accelerations is validated using accelerometer measurements from the GRACE-FO C satellite. Our experimental results demonstrate that decision tree achieves an accuracy of 87.51% in distinguishing different categories of space objects based on combinations of RCS size and object type. This indicates that the proposed approach has significant potential for improving classification in SSA systems.