Research On Vehicle Sparse Trajectory Similarity In City Traffic

With the rapid growing of GPS equipment and wireless communication technology, the diversity and complexity of vehicles trajectory data in urban traffic have increased. So it’s more and more important to mine and analyze the vehicle trajectory data. The vehicle trajectory is sparse in the observation time:vehicle trajectory points distribute unevenly in time dimension, that is time intervals between trajectory points are not equal and the time spans of different vehicle trajectories vary. While the time spans of some vehicle trajectories are long, the others’are short, and time span of vehicle trajectory is a small fraction of the total observation time.Conventional trajectory similarity methods are not suitable for analyzing the sparse trajectory similarity.This paper focuses on sparse vehicle trajectory similarity in urban traffic, analyses their characteristics and the trajectory similarity methods suitable for them.The first chapter introduces the research background and significance, studies the characteristics of the sparse trajectory,and introduces the application scenarios of sparse vehicle trajectory in urban traffic and its research value. The second chapter introduces related work on trajectory similarity,and analyzes the shortcomings of existing methods on sparse vehicle trajectory. The third chapter proposes the thought of key point.discusses the appropriate method of choosing key points in sparse vehicle trajectory and gives the algorithm to find key points. The fourth chapter proposes the thought of time slicing,discusses the effective way to do time slicing for sparse trajectory and gives the algorithm to do time slicing. Based on the thought of key point proposed in the third chapter and the thought of time slicing proposed in the fourth chapter, the fifth chapter proposes a sparse trajectory similarity measurement and a similarity computation algorithm STS(Sparse Trajectory Similarity Computation). Experimental results based on real datasets confirm the effectiveness and efficiency of the proposed algorithm.In this paper, the main work includes:(1)we study the characteristics of sparse vehicle trajectory in urban traffic and discuss the application scenarios and research value;(2)we study the operation effects of conventional trajectory similarity methods on sparse trajectory;(3)we propose the thought of key point and time slicing, and proposes a sparse trajectory similarity measurement and a similarity computation algorithm STS;(4) Experimental results based on real datasets confirm the effectiveness and efficiency of the proposed algorithm.