Computing similarity of coarse and irregular trajectories using space-time prisms

Increasing volumes of trajectory data require analysis methods which go beyond the visual. Methods for computing trajectory analysis typically assume linear interpolation between quasi-regular sampling points. This assumption, however, is often not realistic, and can lead to a meaningless analysis for sparsely and/or irregularly sampled data. We propose to use the space-time prism model instead, allowing to represent the influence of speed on possible trajectories within a volume. We give definitions for the similarity of trajectories in this model and describe algorithms for its computation using the Fréchet and the equal time distance.