Analysis And Prediction Of The Vessel’s Behavior Of Entering And Leaving Port

In recent years,with the economic recovery,countries have paid more attention to the development of maritime traffic,the complexity of ship traffic carried by ports as maritime transportation hub nodes has increased,and the navigation risks have increased,and the efficiency and safety of ship navigation in and out of ports are important guarantees for the smooth operation of ports and the smooth operation of maritime traffic.At the same time,with the gradual improvement of technology and coverage applications of marine traffic communication equipment,ship AIS data has become a cheap but important data resource,and the massive ship dynamic track information provided by it provides the possibility for the analysis of ship navigation process,which is to realize An important method for adding value to AIS data.Therefore,in order to enrich the means of maritime data utilization,better grasp the navigation rules of ships in port waters,and improve the ability of maritime situation awareness and safety supervision,this thesis takes the behavior of ships entering and leaving the port as the research object,and uses data mining,machine learning and other methods to analyze the behavior of ships entering and leaving the port.The main research works of this thesis are as follows:(1)Aiming at AIS data quality and the mining and analysis requirements of ship behavior,an AIS data preprocessing process and related algorithms are designed.As a commonly used data source in marine traffic analysis research,the data quality of AIS directly affects the reliability of mining analysis conclusions,so this thesis summarizes the common quality problems in AIS data,and puts forward the corresponding quality problem detection and cleaning methods,and at the same time considers the coupling influence between different data quality problems to design the process of AIS data cleaning,and verifies the rationality of the method through experiments.(2)In order to finely analyze the spatiotemporal characteristics of ship entering and leaving the port,a statistical analysis method based on spatial grid is proposed,and a lightweight analysis software for ship inbound and outbound traffic characteristics is designed.By dividing the spatial grid of the inbound and outbound channels,the statistical analysis methods of the number of ships,draft,ship speed and other characteristics in the spatial grid of the waterway are given,and a lightweight statistical analysis tool for the traffic characteristics of ships entering and leaving the port is further designed which integrated traffic analysis algorithm based on Py Qt5 framework.(3)In order to effectively mine the behavior pattern of ships entering and leaving port,a typical route extraction algorithm for ships entering and leaving port is proposed.The OD point spatial characteristics and COG characteristics of ship entry and exit trajectories are used to identify the entry and exit trajectories of different types of clusters through hierarchical clustering model,then the typical route patterns are quantified by spatial distribution and spatial density distribution of trajectories with the method of spatial rasterization which provides a pattern data for ship trajectory prediction.(4)In order to apply the navigation pattern information in ship trajectory prediction task,a ship trajectory prediction algorithm based on the improved Seq2 Seq network is proposed.The multi-feature encoders fuse the ship motion status data,spatial position information and historical navigation pattern information of ships entering and leaving the port,then a decoder is applied to predict the longitude and latitude of the ship’s trajectory,the trajectory data of ships entering and leaving the port of Qingdao Port is used to verify the model,which proves that the ship navigation space information and historical navigation pattern information can improve the performance of the ship trajectory prediction task.