Years ago when a car pulled up at a gate, it would take the work of a gatekeeper, and his assistant to check the identity of the visitor, phone the office to verify they were expected, and then open the gate for them to enter. All done in perhaps 5 minutes using the skills of only 3 or 4 people. These days the car can arrive at the gate, slowing gently, and the driver sees the gates automatically open, allowing the car in. The system that allows this is Automatic Number Plate Recognition. How does this system work?
Think of the technology behind this as a chain of links.
– Image capture
– Image processing
– Processing the required signal
– Sending the signal to a processing server
– Comparison with a database
– Database outputs decision
A feed could be taken from an existing camera system but better results come from dedicated cameras. Most (except for some law enforcement agencies) will be fixed and should be positioned to get the best sight of passing vehicles.
To enable operation in darkness, infra-red detectors may be used with suitable infra-red illuminators to enable them.
Systems using multiple cameras can be used to get both, approaching and departing vehicles.
Police, customs and border enforcement agencies may require cameras fitted to vehicles for mobile number capture.
The desired outcome of the image is the unique identifier given by the registration plates.
Machine learning and artificial intelligence techniques are used to find and read the number plates and convert that to a (usually) text output. This can be sent elsewhere for onward processing. GPS (Global Positioning System) data and time/date information can be added to increase evidential strength. A clear visual image can also be attached to allow subsequent human verification and auditing.
Optical character recognition is a process whereby a visual image can be interpreted to yield a text type output from the area identified as the number plate. It is used as part of this processing.
At the collection point, all the elements from the cameras are processed to deliver the composite token of identity for each vehicle that passes. This has to be done quickly, since in a busy location there could be many vehicles passing, each to be processed. It also has to be done reliably, reducing the possibility of failure to achieve the token, or wrong processing which could give a false outcome.
In most instances, the decision making from the unit of information is done at a remote location. Reliable and secure communication has to be established to allow for this.
Checking the Database
A core database is the central part of decision-making for an automatic number plate recognition system. This will contain an up-to-date list of “alloweds” and “forbiddens”. The unit of information from each vehicle must be compared with the database to allow a robust decision to be made. Obviously, for law enforcing agencies, this will require high evidential security. For smaller scale entry control, or parking management systems, as it may be installed by a private operator, less rigorous standards may be allowable.
Delivering the Decision
The decision from the database could go to a range of possible destinations. This could be a warning beep in the patrol car, triggering a gate or generating a payment demand for parking.