Our main motivation stems from a desire to improve our city's existing public transport systems by integrating innovative technologies into the ticket gates found in bus stations. The current bus ticket payment process is prone to security breaches: if a passenger's magnetic card is stolen, the intruder can still use the card until it is manually blocked by a system administrator. Also, the cameras present in bus stations are currently used for security reasons only: the camera output could provide valuable information for crowd management and logistics purposes.
With this in mind, we intend to implement facial recognition technologies to facilitate the process of blocking magnetic cards; and also, by analyzing the camera feed, we want to make use of algorithms that can estimate the amount of people inside a tube station at a given moment in time so that this information could be processed by a logistics central, where buses would be dynamically allocated to regions with higher concentrations of passengers.
Two different Web interfaces will be implemented: one for Admins and one for Regular Users. In the admin point of view, the last image processed by the crowd counting algorithm will be displayed along with a history of the amount of people estimated to be inside the station at the time the picture was taken and also provide information about users that engaged in suspicious activity. This will provide the system administrators with valuable information about the concentration of people inside the tube station (potentially, the processed data could be sent to a logistics central) and aid them in deciding whether an user should be blocked. Regular users, on the other hand, can create and access their profile, insert credits to pay for tickets and see their payment history.
The main function of the mobile app is to provide ease of access to the ticket payment system, so that the passenger doesn't have to carry their magnetic card around. After logging into the application, the user will be directed to the NFC screen, where they will then be able to approximate their smartphone to the ticket gate, pay for the bus ticket and be granted access into the tube station. Inside the mobile application, users will also have access to some of the content displayed in the Web application.
Our cloud services will be responsible for hosting the NodeJS application, PostgreSQL database, VueJS webpage and image processing with Computer Vision Modules, their network connections and WAF (VPC*) configurations, with scalable-ready imaging. We chose the cloud architecture so that our users could rely on its security and accessibility anywhere they go. This also provides scalability for the project: in order to manage the heavy load of image processing, ready to be load balanced and horizontally scaled, the cloud architecture provides the best solution: An algorithm that locates different kind of objects and its respective number, that will be used to measure the number of people inside the tube station. Also Computer vision Marketplace for Cloud services provide an extensive library for facial recognition.
The embedded system will consist of a micro-controller that will be able to read an NFC sensor, enabling users to pay for the ticket with their smartphone or NFC cards. It must also be able to receive images from two cameras, one to perform facial recognition of the user for fraud prevention purposes and another to count the number of people inside the tube station. It will control the turnstile, activating a servomotor that will lock or unlock its mechanism, in addition to reading an infrared sensor that identifies the moment the turnstile turns, being able to lock it again. It will have an LCD display to show the system status to the user, a red and a green led to indicate if the turnstile is locked or not and a buzzer to signal the turnstile unlocking.
The system will communicate with the cloud server, which will be in charge of processing the camera images and storing the users data.
The base station will consist of a few parts: