A Spatiotemporal Trajectory Similarity Measurement Based on Error Ellipses and Stability
Kailai Zhou, Qinglei Meng · 2023
The measurement of trajectory similarity plays a crucial role in the processes of trajectory retrieval, classification, mining, and other trajectory analysis tasks, and it finds widespread application in trajectory data. Existing similarity measurement methods have mostly been developed under the assumption of good data quality, with a predominant focus on spatial aspects while seldom considering both spatial and temporal dimensions simultaneously. An essential challenge related to temporal considerations is dealing with trajectories that have different sampling rates and asynchronous sampling, both of which introduce a degree of uncertainty. To address these issues, this paper presents a novel method for similarity measurement, considering uncertainty, based on error ellipses, termed Spatio-Temporal Uncertain Trajectory Similarity Measurement (STUSM). Experimental comparisons were conducted using real trajectory data and related work. The results indicate that the proposed approach exhibits enhanced robustness when dealing with various challenges such as different sampling rates, asynchronous sampling, and outliers.