Study On Vessel Trajectory Prediction Model Based On AIS Data

With the rapid development of shipping industry and the rapid growth of traffic volume,there are many problems in the water area,such as channel congestion and navigation accidents,which cause huge economic losses.The realization of accurate ship track prediction can not only discover the abnormal track of the ship in real time,but also effectively prevent the landing,collision and other accidents of the ship,and provide technical support for collision avoidance and route planning decision-making of the ship,which is conducive to the maritime traffic supervision.In order to strengthen the control and surveillance of maritime traffic,AIS(Automatic Identification System)came into being.Compared with the traditional radar equipment,AIS equipment has the advantages of high positioning accuracy,less affected by terrain and weather,and can provide rich data sources for ship track prediction.The work is mainly carried out in the following aspects:Aiming at the problem of large order of magnitude and redundant data in the original AIS data,the AIS data is preprocessed.First of all,AIS data should be collected,cleaned and missing values processed.Secondly,in order to ensure the accuracy of data processing,a linear fitting method is selected to interpolate the AIS data with equal time intervals to restore the original trajectory data to the greatest extent.Finally,through comparison,the appropriate ship track prediction model is selected.According to the structural characteristics of ship trajectories,a DBSCAN clustering algorithm based on trajectory segment is proposed to cluster ship trajectories.Firstly,the ship AIS data is used to cluster the routes,and the ship trajectory division model is established.Secondly,the similarity measure algorithm between tracks is optimized to improve the accuracy of clustering algorithm.Finally,the ship track data set is extracted from the discrete original AIS data,which provides the data basis for the track prediction based on AIS data.According to the characteristics of time series in AIS data,in order to achieve the purpose of accurate prediction of ship position,a long-term and short-term memory prediction model based on RNN-LSTM is proposed.Firstly,the influence of various parameters on the prediction model is analyzed.Secondly,in order to better verify the accuracy of the prediction results,the BP neural network model and LSTM model are trained respectively,and the prediction results are compared with the actual track.Finally,according to the prediction results,it is concluded that under the requirements of accuracy and performance,the ship track prediction model based on LSTM can achieve excellent prediction results.