Multi-factor generalization of trajectory data

Lei Zhang,Guang Yang,Zhichao Wang

DOI: https://doi.org/10.3969/j.issn.1001-0505.2012.S2.009

2012-01-01

Abstract:In order to solve the problem that most of current trajectory data generalization methods only consider the spatial characteristic and ignore other trajectory features, thus resulting in information loss, a method of multiple-factor generalization of trajectory data (TRMGEN) is proposed, which considers trajectory features, such as spatial location, time, velocity, direction angle and acceleration, etc. First, the DBSCAN algorithm is improved by adding factors into distance constraint when clustering the points. Then, the representative points of clusters are used in the region division, and the regions are connected to generate the generalized trajectory. As the trajectory multi-factor generalization lacks unified measure standard, from the aspects of information loss and the similarity with original trajectory, a method for measuring generalization effect is presented. Experimental results on the Hurricane dataset show that the TRMGEN can greatly reduce the information loss of the generalized trajectory, and the cluster result of the generalized trajectory is closer to the original trajectory.