Research On Berthing Behavior Of Vessels Based On AIS Data

Port water traffic flow and navigation environment are more complex,berthing operation is more difficult,and with the increase of vessel traffic in the port,the probability of maritime traffic accidents increases.Therefore,to ensure the safety of vessel berthing is an important issue for ship pilots and maritime regulatory authorities.With the automatic identification system is widely used in all kinds of vessels,AIS data including vessel navigation information grows explosively,which has a high research value.Therefore,based on the AIS historical data of Tianjin port,this paper studies the vessel berthing behavior as follows:(1)Aiming at MMSI,vessel position,COG and SOG in AIS data,this paper puts forward data anomaly detection and correction model,establishes AIS data cleaning and storage framework,and provides guarantee for berthing feature analysis based on AIS data.(2)In order to distinguish berth information of berthing track accurately,this paper proposes a method of vessel berthing trajectory classification based on berth point clustering.Based on DBSCAN algorithm,the berthing points are clustered and the vessel trajectories are matched according to the berth point clusters.This method can distinguish the berthing points and the berthing trajectories accurately.Compared with the real berth points,the accuracy of clustering results is verified.(3)When analyzing the characteristics of vessel berthing behavior,based on the obtained vessel trajectory clusters,the vessels of different lengths are classified and analyzed.The vessel arrival distance model is designed according to the SOG characteristics,and the SOG distribution law is obtained by the nonlinear fitting method based on the least square method.According to the COG feature,the COG is divided into 8 azimuth intervals to obtain the COG feature distribution.Besides,the threshold statistical algorithm is designed,which can analyze the characteristic distribution of the vessels passing through the threshold and the vessel flow in each period.(4)Based on the bidirectional GRU network model,this paper proposes a method to predict the vessel's berthing trajectory,constructs the characteristics of berthing trajectory based on AIS data,trains the model by using the berthing trajectory of containers in port waters,and forecasts the future vessel's navigation trajectory.The results show that compared with the LSTM and GRU models,the bidirectional GRU network prediction model has smaller error and more accurate.