In order to solve the problems that dimensionality is high and feature extraction from flight trajectory data is inaccurate at the terminal area, this paper proposes a trajectory pattern recognition method based on LSTM-DAE spectral clustering. Firstly, the paper plans to achieve dimensionality reduction and to extract feature from the processed trajectory dataset by the LSTM-DAE network, and then proceed to even more accurately capture the nonlinear features of trajectory. Secondly, spectral clustering is employed by using the extracted trajectory features to complete pattern partitioning. Finally, an example analysis is conducted on the entry flight trajectory data at Tianjin Binhai Airport. The experiment shows that this method can accurately cluster high-dimensional flight trajectories after extraction, and can be divided into six categories of trajectory clusters, achieving still higher clustering quality. And the method can provide support for effectively identifying flight trajectory pattern features at the terminal area.