Trajectory classification based on Hidden Markov Models

This paper presents a method for statistical modeling and classifi-cation of motion trajectories using Hidden Markov Models. Mass recordings from visual surveillance are processed to extract objects trajectories. Hidden Markov Models of classes of behaviour are created upon some annotated trajectories. In this way, information about complex object behaviour of objects can be discovered. Additionally, an experiment shows the successful application of Hidden Markov Models on trajectories of people in an underground station in Roma. Finally, a comparison of efficiency on different data sets, is discussed