Similarity Measurement of Spatiotemporal Trajectories Considering Semantic Features

Abstract

To solve the difficulties that considering the spatiotemporal and semantic features simultaneously in the similarity measurement of spatiotemporal trajectories, a multi-dimensional semantic matrix of spatiotemporal trajectory is constructed, and a unified quantitative semantic space is established by combining the singular value decomposition. The spatiotemporal trajectory is mapped into a feature semantic vector and a low rank feature matrix, containing the essential characteristics of trajectory, and the data dimensional reduction and denoising are realized after the decomposition process. Comparative experiments on real dataset demonstrate that the proposed method can accurately describe the essential characteristics of spatiotemporal trajectories on the basis of geospatial level, and effectively measure both semantic and geographic similarities among trajectories.