As India is a country with a huge crowd, travelling via air becomes very tiresome because of the processes before the air-travel. Since no passenger likes being interrogated multiple times by the security personnel, We have come up with an idea which makes the work of the security sector as well as the passenger easy and hassle-free. We wanted to introduce such a system which makes passenger movement inside the airport free and also making the work of the security easy.
What it does
Our system keeps track of every passenger within the airports using a Two Node Human Face Tracking system[TNHFT] that we developed in the overnight hackathon which runs on a central artificial neural network of the airline. This neural network which is floating on a central server of Vistara tags each passenger's face who is traveling through Vistara with his ticket_id and will be able to track as well as recognize the passenger's status, i.e, if he's boarded on the flight or checked in, arrived at the destination airport etc. And our TNHFT system can also play a major role in the airport security sector by accessing the ticket_id to track the person of interest by logging in his status, and hence easing the process of manually finding the person in a camera stream.
How we built it
We have built a system consisting of a Mobile Application which can scan the ticket and grab the ticket_id and link it to the respective passenger holding it and this data is fed into the Artificial Neural Network which would be floating on a central server. Since this is a two node system, we have an initial camera connected to the CNN which links the ticket_id of the passenger and starts detecting his face and logs it into a category labelled by the ticket_id. There's going to be another camera at the destination which will confirm his arrival at that check-point. We've used OpenCV to detect the faces, Keras using Tensorflow as backend, Picamera for raspberry pi to start capturing images. We've used Python to build the entire system.
Challenges we ran into
Building the neural network. Building the server, since we require real time facial recognitions which is possible over a powerful GPU
Accomplishments that we're proud of
Successfully training our neural net and logging the predictions Face detection using OpenCV
What we learned
Face detection using OpenCV which saves frames of the face into a folder.
What's next for Computer vision based security systems
Upcoming version would remove Ticket_id and include a national UID like Aadhar, thereby gradually changing the legacy trend of passengers carrying multiple verification documents to the airport.