Securing harbors by modeling and classifying ship behavior
Résumé
Since 2001, Works in the field of security have been considerably growing. All over the word, Public places such as markets, Car parks, Hotels, Metro and train stations are permanently threatened by terrorism. For this reason, Researchers work constantly on finding ways to ensure security. In this paper, And as researchers working on a French project aiming to protect harbor facilities and people from threatening events, We focus on ship behavior while evolving in the harbor. We propose using the probabilistic approach Hidden Markov Models HMM because of its promising performance in the field of behavior learning and recognition. The target of our contribution is to be able to statistically classify each ship's behavior as it enters into harbor and determine whether its behavior is common or unusual. Our idea was to use a map of the port as well as ship information (positions, Etc.) previously collected from sensors to construct a set of all possible behavior patterns in the harbor. Then, This set would be used on-line to recognize new ships entering the harbor. Some criteria were set to consider a ship's behavior as unknown or unusual. The harbor map made it possible to initialize HMM models of ship behavior, Then the well-known Baum-Welch algorithm was chosen to learn models from ships' trajectories obtained from port and finally the forward algorithm was used to classify and recognize each new type of ship behavior.