Real-time tracking using camera networks and multi-modal sensing with the Foglets Framework

Georgia Tech police department is using a proprietary system to manage over 700 cameras on campus. The current approach requires administrators to click on each camera pin on the map and watch the stream to find out useful information, or wade through backup camera videos to track incidents. This process requires a lot of human labor and is not scalable. We use fog computing infrastructure to provide real-time object tracking, with car tracking across space and time as a concrete use-case. 

Administrators only need to input the description of the car, such as license plate number, color, model, or a picture, and the system will return the trace of the car that is covered by the campus cameras if found. Using technologies like sensor fusion, our camera system can be turned into a general multi-sensor system that can be used to track cars, people, detect anomalies and more by universities, companies, and governments. For example, we can use sensors like LPR (license plate reader) and TPMS (tire pressure monitoring system) to improve the accuracy of our camera network to do the car tracking.