Abstract: The trajectory pattern recognition can further support track prediction, airspace structure optimization and air traffic situation monitoring. Firstly, a series of trajectory point sets are formed by screening and processing the initially obtained automatic dependent surveillance-broadcast (ADS-B) data. Secondly, the similarity degree between trajectory data composed of trajectory points is calculated. The training set and test set are selected according to a certain ratio; then, the training set is used for learning and training, and a self-organizing neural network (SONN) is established to obtain a competition unit. Finally, the test set is input and identified by the resulting neural network to analyze its correctness. Simulation results show that the algorithm can better identify the trajectory in the airspace of terminal area.

Key words: terminal airspace, trajectory classification, pattern recognition, self-organizing neural network (SONN)