M-STCP: an online ship trajectory cleaning and prediction algorithm using matrix neural networks

Abstract

Accurate prediction of ship trajectories is crucial to guarantee the safety of maritime navigation. In this paper, a matrix neural network-based online ship track cleaning and prediction algorithm called M-STCP is suggested to forecast ship tracks. Firstly, the GPS-provided historical ship trajectory data is cleaned, and the data cleaning process is finished using the anomaly point algorithm. Secondly, the trajectory is input into the matrix neural network for training and prediction, and the algorithm is improved by using Kalman filtering, which reduces the influence of noise on the prediction results and improves the prediction accuracy. In the end, the effectiveness of the method is verified using real GPS trajectory data, and compared with the GRU model and long-short-term memory networks. The M-STCP method can improve the prediction accuracy of ship trajectory to 89.44%, which is 5.17% higher than LSTM and 1.82% higher than GRU, effectively improving the prediction accuracy and time efficiency.