Due to problems such as equipment failure, transmission delay, and signal loss, the original AIS (Automatic Identification System) data generally need to be preprocessed before further analysis. In this paper, we propose a novel approach for vessel trajectory reconstruction combining deep learning networks Long Short Term Memory (LSTM) with variable modeling, and a trajectory data unsupervised learning architecture is conducted, including 1) abnormal trajectory data identification and cleaning; 2) vessel navigational states identification; 3) vessel trajectory reconstruction. In this study, we have investigated and proposed a novel method which use LSTM to reconstruct the longitude and latitude of vessel trajectory data respectively, and provide effective trajectory data for subsequent collision avoidance, vessel type analysis, risk evaluation, trajectory prediction, route planning and other research. More importantly, field measured vessel data is collected by a real-time detection system to verify the proposed method.
展开▼