Figure 1

Trajectory Generalization process. (a) Original trajectories from Milan GPS dataset; (b) simplification of trajectory points by considering only relevant events (start, end, turns); (c) relevant points are clustered by spatial proximity to reveal zones with high frequency of movement events; (d) from the centroids of each cluster it is derived a voronoi tessellation of the territory; (e) original trajectories are generalized as sequence of traversed cells of the previous tessellation.