In this paper, we study the problem of finding a class of spatio-temporal patterns called (m, k)-flock patterns (Gudmundsson and van Kreveld, Proc. ACM GIS'06; Benkert, Gudmundsson, Hubner, Wolle, Computational Geometry, 41:11, 2008), which represent a groups of moving objects close each other within width at most r under L_∞-norm in a given time segment of length at least k, in a collection of 2-dimensional trajectories. For max-width r > 0, min-length k, and a collection S-of n trajectories of legnth T, the proposed algorithm finds all length-maximal (m, k) flock patterns in an input collection of trajectory data in 0(pnT~2) delay and 0(p~2) space, p = |X| is the size of ID set being enumerated. We also present a practical improvement using geometric indexes.
展开▼