DAISTIN: A Data-Driven AIS Trajectory Interpolation Method

Abstract

The Automatic Identification System (AIS) provides global vessel positioning data used in a variety of maritime applications. However, AIS suffers from transmission signal gaps, which causes vessels to disappear from AIS records for prolonged periods and poses a major challenge for the use of AIS data. In this paper, we propose a novel Data-driven AIS Trajectory INterpolation method (DAISTIN) to address AIS signal gaps. DAISTIN first makes use of massive raw AIS data to delicately construct a graph that well represents vessel movements. Next, given a gap between two locations A and B in an AIS trajectory, DAISTIN searches the graph for the shortest path from A to B and uses the path to interpolate the vessel’s whereabouts in between. To cope with large amounts of AIS data, we design a geometric sampling method for DAISTIN to select representative AIS data points for the graph construction. Finally, we design a postprocessing step for DAISTIN to fine-tune the quality of interpolated results. We conduct extensive experiments to compare DAISTIN with selected existing methods. The results verify the superiority of DAISTIN in terms of multiple performance metrics.