Ship Trajectory Clustering Based on Image Feature Distance
In order to further optimize the management of maritime traffic,a ship trajectory clustering algorithm based on dis-tance metrics of trajectory image features is proposed.This algorithm aims to address the problems of difficult setting of weight pa-rameters and long running time for traditional ship trajectory clustering algorithms based on multiple dimensional attributes.The al-gorithm utilizes Automatic Identification System(AIS)data to draw trajectory images based on the position,speed,and course of trajectory points.The trajectory image features are extracted via a deep residual network trained on large-scale image data.The fea-ture dimensionality is reduced via principal component analysis.The distance measure between trajectories is based on the Euclide-an distance of feature vectors.The density-based noise-tolerant clustering algorithm(DBSCAN)is employed to cluster the reduced ship trajectory image features.Experiment results show that the proposed algorithm can effectively cluster the trajectories while re-ducing the running time.The characteristics of the ship traffic flow reflected by the trajectory clusters are consistent with the actual situation.
ship trajectory clusteringship trajectory distance measureDBSCANcharacteristics of the ship traffic flow
国家科技期刊平台
NSTL
万方数据
