Discovering Companion Vehicles from Live Streaming Traffic Data

摘要

Companions of moving objects are object groups that move together in a period of time. To quickly identify companion vehicles from a special kind of streaming traffic data, called Automatic Number Plate Recognition (ANPR) data, this paper proposes an approach to discover companion vehicles. Compared to related approaches, we transform the companion discovery into a frequent sequence-mining problem. We make several improvements on top of a recent frequent sequence-mining algorithm, called SeqStream, to handle customized time constraints among sequence elements when discovering traveling companions. We also use pseudo projection technique to improve the performance of our algorithm. Finally, extensive experiments are done using a real dataset to show efficiency and effectiveness of our approach.