An application of convolutional neural network to derive vessel movement patterns

The rising need for highly effective maritime transportation, the demand for a higher safety level for protecting maritime environment and the intelligent ships’ development have been the driving forces towards Maritime Autonomous Surface ships (MASSs). Navigational devices such as the Automatic Identification System (AIS), Long Range Ship Radar (LRSR), high precision camera and Electronic Charts make it possible for MASSs to derive comprehensive maritime Situational Awareness (SA) by using the information collected by these devices. The vessel movement pattern extraction from these devices is one of the important parts for SA. In this paper, our research proposes one method called vessel movement pattern extraction from trajectory images using AIS data to classify the vessel movement patterns (static, cruise and manoeuvring). Firstly, we rebuild the vessel’s trajectories with historical AIS data and label the AIS data. In order for Convolutional Neural Network (CNN) to easily handle the raw AIS trajectories, our method converts raw AIS trajectories into an image data structure while keeping the vessel movement pattern information. And then, the CNN method is used to extract the vessel motion patterns. Finally, we demonstrate the effectiveness of the proposed method through several experiments using AIS historical dataset.