Date: June 2016 | Client: Airport of Puglia
The solution implemented at Aeroporti di Puglia” (http://portal.aeroportidipuglia.it) allows to anonymously and passively capture passengers’ facial templates in the airport area and track them through their journey.
The main scope of the solution implemented is to:
“Trace passengers within the airport area by tracking multiple faces simultaneously, instantly compare images coming from video streams and then identify a suspected person by real-time comparison to a watch list databases”
An explicit criterion for the Airport Security Manager was to guarantee that installation of a new equipment does not interfere with the normal activity of the airport.
The integration involves the installation of video capture devices, configured and installed according to application contexts (1:n Verification type). This scenario provided Reco the opportunity to deduce complex operational contexts and identify performance indicators of the solution.
1. Passengers enrollment at entrance
Passenger's face is automatically detected and captured by the system at the entrance to the airport. The relative image is encoded by creating a numerical representation and stored in a database with a timestamp. The physical image instead, is automatically discarded after a couple of hours (as we assume that passenger’s presence inside the airport areas is temporarily over).
Anyhow, the entire detection and enrollment process occurs passively and is not invasive - the passenger does not need to stop or interact with the camera, so the airport operational flow is not impacted.
2. Matching process inside the airport
Inside the airport areas, several Reco nodes have been installed that are able to capture a passenger's facial image. Each image is encoded into a template and used to trigger a search process on the database for potential matches.
Once a match is found, a notification to a Control room is submitted with regards to the recognition occurred. In specific analytical cases the timestamps can be used to determine how long it has taken the passenger to go from one node to another.