Abstract：In this paper, a trajectory anomaly detection method for aircraft in terminal airspace surrounding an airport is proposed. This method assists aircraft condition monitoring system (ACMS) for safety analysis, as an effective and efficient way for anomaly detection. Considering the characters of aircraft, a trajectory reconstruction method based on ADS-B data is presented. Then, the mixture density network is utilized to extract the standard trend. Furthermore, make accurate inferences about trajectories. Finally, the trajectories are classified as either normal or abnormal based on the probability assessment method. Extensive tests on real flight data show that proposes models behave well in anomaly detection and are suitable for expanding flight data.

CICTP