Inferring Semantically Enriched Representative Trajectories

In the analysis and visualisation of clustered spatial trajectories, the computation of a representative trajectory for a given cluster of data trajectories plays an important role. Usually, such a representative trajectory is computed based upon the data trajectories' spatial characteristics only, e. g., as an average, median, or central trajectory. However, in many cases, the input data is enriched by various types of semantic information which may document characteristics of the trajectories as well. We present an approach to inferring representative trajectories for a given cluster of trajectories. Our approach constructs an extended finite state machine describing the spatial and non-spatial properties of the data trajectories in a given cluster. This extended finite state machine then can be used to generate a representative trajectory exhibiting characteristic changes in spatial and non-spatial properties. The extended finite state machine constructed is annotated with these changes, hence enabling domain experts to further analyse and assess the constructed representative trajectory.