A Method for Improving Two-line Element Outlier Detection Based on a Consistency Check

As the most complete source of orbital element information available to the public, the NORAD two-line element sets (TLEs) are used in a wide variety of orbit propagation tasks. Unfortunately, there is no error information provided for the TLEs. Due to orbit manoeuvres, errors introduced during the TLE generation and unmodelled perturbations, there are inevitable outliers in the TLEs, which have a large deteriorative impact on orbit determination and propagation. Most of current methods identify outliers using a three-sigma rule or a Mahalanobis distance-based detection method. However, in these methods the perturbation characteristics of space objects in different altitudes are not taken into account. This study presents a method for detecting outliers in the TLEs of space objects based on a consistency check on pair-wise differential residuals of a series of TLEs. A filter based on the principle of locally weighted regression is applied on the pair-wise differential residuals to investigate their underlying structure. The detection threshold is then determined by the variance in a moving window running over the filtered residuals. Satellites from different altitudes with known manoeuvre histories were selected as examples to demonstrate the effectiveness of the proposed method. Our results show that reliable detection of the manoeuvre events can be achieved. The differences in the characteristics of TLE outliers between satellites and debris objects are also analysed to facilitate the application of the satellitebased method to debris to identify erroneous TLEs. It is expected that the improvement made by the new method will contribute to more robust orbit propagation and conjunction analysis that are based on TLEs.