Dwell Regions: Generalized Stay Regions for Streaming and Archival Trajectory Data

Abstract

A region ℛ is a dwell region for a moving objectOif, given a threshold distancerqand duration τq, every point of ℛ remains within distancerqfromOfor at least time τq. Points within ℛ are likely to be of interest toO, so identification of dwell regions has applications such as monitoring and surveillance. We first present a logarithmic-time online algorithm to find dwell regions in an incoming stream of object positions. Our method maintains the upper and lower bounds for the radius of the smallest circle enclosing the object positions, thereby greatly reducing the number of trajectory points needed to evaluate the query. It approximates the radius of the smallest circle enclosing a given subtrajectory within an arbitrarily small user-defined factor and is also able to efficiently answer decision queries asking whether or not a dwell region exists. For the offline version of the dwell region problem, we first extend our online approach to develop the ρ-Index, which indexes subtrajectories using query radius ranges. We then refine this approach to obtain the τ-Index, which indexes subtrajectories using both query radius ranges and dwell durations. Our experiments using both real-world and synthetic datasets show that the online approach can scale up to hundreds of thousands of moving objects. For archived trajectories, our indexing approaches speed up queries by many orders of magnitude.