Detection And Early Warning Method Of Ship Abnormal Behavior Based On Massive AIS Data

With the wide application of AIS,maritime regulatory authorities can obtain massive AIS data,and ship behavior detection has entered the era of big data.However,due to various reasons,there are many problems with AIS data such as data loss and a lot of errors,the current quality of AIS data can not meet the needs of maritime supervision and service.It is of great significance to improve the availability of spatiotemporal trajectory data,automatically detect the abnormal behavior of ships in ports and sensitive waters and realize early warning,which is very important to ensure the safety of water traffic.Therefore,focusing on the topic of ship abnormal behavior detection,this paper carries out research work from abnormal AIS data discrimination,abnormal behavior detection of ships in port waters and sensitive waters,the specific research contents are as follows:In the aspect of abnormal AIS data discrimination,combined with the characteristics of adjacent AIS data in a period of time,an abnormal AIS data discrimination model based on BP neural network was constructed.By dynamically adjusting the learning rate of the neural network,the learning efficiency of the network model was improved.The model has been trained and verified with AIS data,and the accuracy rate reached 95.16%.At the same time,the influence of AIS data segment length and hidden layer node number on accuracy was analyzed through experiments.The experimental results showed that unreasonable data segment length and the number of hidden layer nodes would reduce the accuracy of the model.When the length of data segment set to 4 and the number of hidden layer nodes set to 6,the accuracy of the model reached the highest.In terms of ship abnormal behavior detection in port waters,according to the characteristics of ships entering and leaving the port,the abnormal behaviors are divided into three categories:abnormal ship spacing,abnormal ship entry and departure,and abnormal ship trajectory.Aiming at the abnormal distance between ships,an abnormal behavior detection method based on ship domain was proposed.Based on the ship AIS data,using statistical method,the ship domain model was constructed and verified by historical data.The detection of single ship domain intrusion,continuous ship domain intrusion and regional ship domain intrusion was realized.Aiming at the abnormal behavior of ship entering and leaving port,according to the motion characteristics of ship's position,speed and course,this paper put forward an efficient algorithm which could accurately judge the ship's entering and leaving port and berthing,and it was verified by the actual data.Through the prior global port geographic information data,combined with AIS message,the algorithm was designed to realize the rapid matching between ships and ports to meet the real-time needs of online early warning.In view of the abnormal trajectory of the ship,on the basis of fully considering the characteristics of ship starting point and terminal point,K-means algorithm was used to cluster ship trajectories firstly.By extracting multiple trajectory feature attributes,the trajectory combination distance was constructed,and the sub trajectories were clustered by DBSCAN.The rules for selecting parameters of DBSCAN algorithm in ship trajectory clustering were set,and the ship sub trajectories clustering and abnormal trajectory identification were realized by selecting different domain radius values.The effectiveness of the method was verified by the actual data.In the aspect of ship abnormal behavior detection in sensitive waters,the core identification factors were selected to construct the analytic hierarchy process(AHP)model.Combined with the characteristics of the research area,the weights of each layer element in the hierarchical model were set by Delphi expert survey method.Based on different abnormal factors,the abnormal behaviors of ships in different sensitive waters were detected,and the abnormal behaviors of ships in the waters near 981 platform,Chunxiao oil and gas field and Huangyan Island were detected.This paper has solved the problems of abnormal AIS data discrimination,abnormal behavior detection of ships in port waters and abnormal behavior detection of ships in sensitive waters.The research results provide the basic conditions for the wider application of AIS,and help to improve the application scope of AIS data,which has important practical application value and practical significance in maritime supervision.