Measuring the Geometric and Semantic Similarity of Space-Time Prisms Using Temporal Signatures

摘要

Well-established techniques exist for measuring the similarity of space-time paths. These measures support clustering and aggregation of space-time paths as well as moving objects database queries based on similar movement patterns or semantics. Little attention has been paid, however, to the analogous problem of measuring space-time prism (STP) similarity, despite comparable applications. This article presents and evaluates a method for measuring STP similarity through dimensionality reduction that leverages their inherent temporal ordering. The technique sweeps an STP along the time axis and derives one-dimensional temporal signatures based on a measured STP property that captures its geometry or semantics. These temporal signatures can be visualized directly as curves. We can also apply existing space-time path similarity measures to these signatures. To demonstrate the feasibility of this approach, we perform two sets of experiments measuring geometric and semantic similarity among STPs and assess the information within these curves using visualization, Frechet distances, and clustering techniques. Results suggest that the temporal signature curves capture meaningful similarities and differences among STPs.