Anomaly Detection from TLE Data

This paper presents a framework for the identification of anomalies in the trajectory of Low Earth Orbit (LEO) satellites. The algorithm can identify a broad class of anomalies from two-line element (TLE) data. The method is based on a combination of time series pre-processing techniques and Long Short Term Memory (LSTM) Autoencoder. The main contributions of this study are utilization of Gaussian derivative filter and demonstrating the ability to detect anomalies on rocket bodies, which have not been previously demonstrated.