Sequential Vessel Trajectory Identification Using Truncated Viterbi Algorithm

Abstract

In this work, we propose a novel classification algorithm that used to classify vessel data points into different trajectories. The algorithm is a truncated version of Viterbi Algorithm. A physical model utilizing the observation information completely is used to simulate the movement of vessels during the period. Distributions of observation noise (also called residuals) are learned from the model. A directed graph is then constructed based on those distributions to portrait the relationship between data points. Truncated Viterbi Algorithm (TVA) is applied to this graph to find the most possible trajectories embedding in the data set. By doing experiments on the maritime domain and Automatic Identification System (AIS) data, we can demonstrate the efficacy of our algorithm.