Our application is targeted to track all vehicles across live camera streams in real time. The result is stored as trajectories of vehicles in a graph database for later queries such as “where did the suspicious vehicle come from and head to?”.
The initial motivation of the project is to facilitate the Georgia Tech Police Department’s procedure of tracking suspicious vehicles, which is largely done manually now. Broadly, given the easy accessibility of cameras, our application could help to improve the city’s public security without acquiring high-cost advanced sensors. Technology: The project makes use of state-of-art computer vision algorithms to do vehicle detection and re-identification from live camera streams. As a systems lab, we explore proper platform services and infrastructures to deploy this application on the edge computing continuum. Edge computing can not only offer low-latency operations but more importantly eliminate the needs of pushing hundreds of live camera streams to Cloud which results in a network burden.
Team Information: Embedded Pervasive Lab (EPL) at Georgia Tech, Atlanta GA