Clustering network-constrained uncertain trajectories
Jingyu Chen, Ping Chen, Qiuyan Huo, XU Xue-zhou · 2011
Low sampling-rate and uncertain features of trajectory data present new challenges to trajectories data mining. This paper proposed a relationship graph-based trajectory clustering algorithm for objects moving on road networks. By constructing an approximate minimum spanning tree of a trajectory, based on the spatial distance of candidate segments, a distance measurement scheme is presented to judge the degree of similarity. The relationship graphical model is adopted to represent the network-constrained trajectory data. A modified RepStream clustering algorithm is proposed to retain the stable relationship information. The experiments show that the clustering algorithm has superior accuracy in low sampling-rate and sampling error trajectories data.