Density based spatio-temporal trajectory clustering algorithm

Abstract

In recent years, with the development of mobile location services and cloud computing technology, the collection and processing of mobile location information has become a reality. The mass database which is composed of mobile location data has promoted the development of the research on mobile location data. It is very important for us to understand the spatial distribution and temporal characteristics of moving patterns and identify the mechanism of motion formation, predict the future development of sports through trajectory clustering analysis. At present, trajectory clustering research mainly focuses on the spatial position changes of moving objects. Temporal constraints in spatial and temporal clustering are generally auxiliary information, but do not really participate in clustering. In this paper, a clustering algorithm for trajectory data based on spatiotemporal pattern is proposed. First, the curve edge detection method is used to extract the trajectory feature points. Then the trajectory is divided into sub track segments according to the trajectory feature points. Finally, the density based clustering algorithm is applied to cluster according to the temporal and spatial similarity between sub trajectories. The Hot spot analysis experimental based on Chengdu taxi GPS track data results show that the similarity measurement based on spatiotemporal features can get better clustering results because of both the spatial and temporal characteristics of the trajectory.