AIS trajectory classification based on IMM data

The importance of the maritime vehicles makes necessary the implementation of systems capable of ensure the safety and security. This paper presents an analysis on Automatic Identification System (AIS) data processed with Interacting Multiple Model (IMM) filter in order to help trajectory data analysis for predictive tasks. The main objective is building a system capable of classifying ships trajectories into different categories as the ship type or the type of activity (fishing, under way with engines, etc.) based on the kinematic and other filter outputs. An automated processing system is implemented to use raw AIS data, preparing and organizing it in order to classify them in ship types and maneuvering state. The appropriate modelling with dynamic models and transition probabilities allow the identification of patterns helpful for trajectory reconstruction and classification. Important aspects as data cleaning, processes parallelization and parameter analysis are dealt on the paper, with results obtained from an available data set.