An Approach to Instant Discovering Companion Vehicles from Live Streaming ANPR Data

An Approach to Instant Discovering Companion Vehicles from Live Streaming ANPR Data

Companions of moving objects are object groups that move together in a period of time. This paper proposes to instantly discover companion vehicles from a special kind of streaming sensor data, called Automatic Number Plate Recognition (ANPR) data. Compared to related approaches, we transform the companion discovery into a frequent sequence mining problem. We make several improvements on top of our previous work, including one scan and tree traversal reduction, to optimize the performance of our previous approach and accelerate the process of discovering companion vehicles. Finally, extensive experiments are done to show efficiency and effectiveness of our approach.