In order to more effectively master the distribution rule of air traffic flow and improve the clustering efficiency and quality of flight trajectories, a clustering method with fast calculation speed, self-identification of original trajectories and elimination of interference trajectories is proposed. First, the resampling method reduces the size of data and arranges the sampling points evenly. Then, the kernel principal component analysis method is used to reduce the dimensionality of the track and increase the difference between different types of points to facilitate clustering. Finally, the DBSCAN algorithm is used to identify and eliminate the interference flight trajectory. Experiments show that this method can eliminate all the disturbing tracks and some of the incomplete tracks that affect clustering. Compared with other clustering methods, the method in this paper simplifies the pre-processing of the flight trajectory before clustering, improves the clustering efficiency and at the same time the clustering effect is more accurate and identify abnormal trajectories.