•  
  •  
 

Abstract

With the rapid development of the global economy, international imports and exports have become increasingly important. Timely and effective tracking and prediction of ship trajectories are important in the face of complex maritime traffic. Based on AIS data, this study establishes a ship trajectory prediction model combining improved particle swarm optimization (IPSO) and a convolutional neural network (CNN) that uses the ship's historical navigation trajectory data to predict the future navigation trajectory. The historical ship trajectory data of the Port of Ningbo-Zhoushan were selected for the experiment, and the IPSO-CNN model experiment results were compared with those of other models. According to the experimental results, the IPSO-CNN model exhibited good performance and high prediction accuracy. The prediction trajectory is more fitting with the real route, and the error is less than 0.001, which is reduced by 57.54% compared with the original CNN model, and the prediction accuracy is higher. Therefore, IPSO-CNN is well-suited for predicting ship trajectories and can achieve accurate prediction. They play important guiding roles in future ship hazard warnings, ship anomaly monitoring, and other applications.

Download
COinS