A novel ship trajectory reconstruction approach using AIS data

Abstract

AIS data plays an increasingly important role in collision avoidance, risk evaluation, and navigation behavior study. However, the raw AIS data contains noise that can result in wrong conclusions. We propose a multi-regime vessel trajectory reconstruction model through three-steps processing, including (i) outliers removal, (ii) ship navigational state estimation and (iii) vessel trajectory fitting. This model allows for vessel trajectory reconstruction in different navigation states, namely hoteling, maneuvering, and normal-speed sailing. The normal-speed navigation trajectory is estimated with a spline model, which can fit any types of the trajectory even with circles. Then, the proposed model is tested and compared with other three popular trajectory reconstruction models based on a large AIS dataset containing the movement of more than 500 ships in Singapore Port. The results show that the proposed model performs significantly better than the linear regression model, polynomial regression model, and weighted regression model. The proposed model can decrease the abnormal rate of speed, acceleration, jerk and ROT (Rate of Turn) from 43.42%, 10.65%, 59.25%, 50.33%–0.00%, 0.00%, 17.28% and 15.81%, respectively. More importantly, the navigational behavior, such as turning operation, could be clearly shown in the trajectory reconstructed by the proposed model.