Trajectory Clustering Based on Trajectory Structure and Longest Common Subsequence

Abstract

Trajectory clustering is an important method for mining valuable information from spatio-temporal data. Longest common subsequence clustering algorithm has advantages in distinguishing the overall trend of the trajectory, but ignores the trajectory structure details. At the same time, trajectory structural similarity clustering algorithm performs recognizing track structural details better than overall trends .Therefore we come up with a trajectory clustering algorithm based on trajectory structure and longest common subsequence in order to combine the advantages of them. The point distance calculation in LCSS is changed to trajectory structural distance which reflects the trajectory details better. The partition algorithm in trajectory structure clustering is improved to reduce the influence of noise points. The simulation results shows that TS-LCSS performs better than TC-SS when we need to recognize trajectories’ over all trends.