ATeDLW: Intelligent Detection of Abnormal Trajectory in Ship Data Service System

In order to improve the intelligent level of ship data service system, it is necessary to detect the abnormal ship trajectories in time. Therefore, we propose an Abnormal Trajectory detection method combining emd-DBSCAN algorithm and LSTM model with Weighted loss function (ATeDLW). Based on the earth mover's distance, the emd-DBSCAN algorithm is proposed and used to cluster ship trajectories, and ship trajectories with distinctive behaviors are identified as abnormal. In order to make full use of the sequence characteristics of trajectory points, the LSTM network is used to train the abnormal ship trajectory detection model, and the weight of abnormal trajectory in the loss function is increased. Real AIS data are used for experiments. The experimental results show that the detection effect of ATeDLW method is better than other methods, and the ATeDLW method can meet the requirements of abnormal ship trajectory detection in different scenarios by adjusting the parameters.