Dependable Clustering Method of Flight Trajectory in Terminal Area Based on Resampling

To master the complex and changeable spatial distribution characteristics of air traffic flow in a terminal area accurately, and to evaluate and optimize arrival and departure procedures effectively, the cluster problem of 3D real flight trajectories in a terminal area was addressed using a resampling technique. A clustering method that has high computation speed, good expandability, and strong dependability was also proposed. First, based on resampling and principal component analysis method, projection of high-dimensional trajectory data to low dimension was implemented based on the premise of maintaining flight characteristics. Then, flight trajectory cluster analysis and outlier trajectory detection models were presented based on the MeanShift method. Finally, the proposed method was verified using real flight trajectory data of terminal areas in order to analyse the effect of every parameter on cluster results. Experimental results show that principal components having a 96.16% accumulative contribution rate can be obtained in 0.004 s. Flight trajectory data can be well approximated by the principal components. Compared with the hierarchical clustering method, the proposed method can obtain more dependable flight trajectory clustering results which correspond to the standard arrival routes. Low similarity trajectories were detected as outliers.