Trajectory Similarity Search with Multi-level Semantics

Abstract

AbstractWith the widespread popularity of intelligent mobile devices, massive trajectory data have been captured by mobile devices. Although trajectory similarity search has been studied for a long time, most existing work merely considers spatial and temporal features or single-level semantic features, thus insufficient to support complex scenarios. Firstly, we define multi-level semantics trajectory to support flexible queries for more scenarios. Secondly, we present a new “spatial + multi-level semantic” trajectory similarity query, and then propose a framework to find k most similar ones from a trajectory database efficiently. Finally, to hasten query processing, we build a multi-layer inverted index for trajectories, design 4 light-weight pruning rules, and propose an adaptive updating method. The thorough experimental results show that our approach works efficiently in extensive and flexible scenarios.KeywordsMulti-level semanticsInverted indexTrajectory similar query