Searching Similar Trajectories Based on Shape

Similarity search in moving object trajectories is a fundamental task in spatio-temporal data mining and analysis. Different from conventional trajectory matching tasks, shape-based trajectory search (STS) aims to find all trajectories that are similar in shape to the query trajectory, which may be judged to be dissimilar based on their coordinates. STS can be useful in various real world applications, such as geography discovery, animal migration, weather forecasting, autonomous driving, etc. However, most of existing trajectory distance functions are designed to compare location-based trajectories and few can be directly applied to STS tasks. In order to match shape-based trajectories, we first convert them to a rotation-and-translation-invariant form. Next, we propose a distance function called shape-based distance (SBD) to calculate the accurate distance between two trajectories, which follows an align-based paradigm. Then, to accelerate STS, we propose a trajectory representation framework based on symbolic representation to support efficient rough match. Finally, extensive experiments on two real-world datasets demonstrate the effectiveness and efficiency of our framework.