printer friendly version

Automatic number plate recognition

The use of number plate recognition technology has expanded dramatically over the last few years. Despite this, the technology remains elusive for some and many players in the security industry are sceptical because in some instances unsatisfactory results and poor system performance have been achieved.

Technology

Automatic number plate recognition (ANPR) is simply the ability to automatically extract a vehicle number plate's characters from an image and then do something with it. In essence it consists of a 'frame grabber' that has the capability to grab an image, find the location of the number in the image and then extract the characters using an optical character recognition (OCR) tool to translate the pixels into numerically readable characters.

Some of the pitfalls that new ANPR technology strives to overcome include blurring of the number plate characters as a result of poor focus or vehicle movement, acute presentation because of poor camera angles and low contrast. The numbers can be obscured because of vehicle headlights or reflections from the sun and many other factors may also influence the clarity of the images.

Applications

While the technology is offered by a number of companies, the proper application is often unclear and begs the question: Once the number has been read, what do we do with it?

The main areas of use are:

Vehicle theft control

The objective of vehicle theft control is to limit the possibility that an unauthorised person is driving out of a secure area with a vehicle. Clearly, knowledge of the vehicle entering does not limit exiting of the vehicle as in normal circumstances this would be legitimate. So it is therefore important to link the driver of the vehicle to the vehicle. This can be achieved with facial recognition of the driver but often the results of such systems are poor, frequently caused by the drivers themselves either on entry or exit and this experience limits the use of this technique. Other biometric links, such as fingerprint reading, require an action and co-operation from the driver. When vehicle owners legitimately allow other drivers to use their vehicle it becomes more complicated, as it requires other drivers' fingerprints to be registered to the car as well.

A practical solution is to issue the driver with a token. While this also requires an action from the driver, it is a relatively simple one - the token, which can be a parking ticket or card, is retained by the driver and carried with him after leaving the vehicle. The vehicle is linked to the token for authorised exit from the protected area. As the transfer of this token to another driver in the vast majority of cases would be with consent, use of the vehicle by another driver could be assumed to be authorised. Loss or theft of the token gives the owner of the token an opportunity to raise an alarm.

The recording of driver faces is mostly used to resolve those situations where the automatic detection of either the token or the number plate failed, the latter usually being due to poor positioning and lighting, loss of number plate or obscured details.

Vehicle access control

While this may share some characteristics with the previous explanation, the system process is different. While vehicle theft prevention may be part of the requirement, access control implies limiting access to pre-authorised vehicles. This requires systems to have lists of people with approved access.

In a residential area this could also include theft control. Linking the driver to a vehicle is more problematic as many drivers are often associated with a single vehicle. Thus multiple tokens or faces need to be associated with a number plate.

An access system could register that a vehicle with a particular number plate is allowed into the premises, while recording driver details only as a precaution. Upon exit driver biometrics (face, fingerprint) or tokens could be used to link any one of the authorised drivers to the vehicle and in this case, unlike the theft control, there is no necessary link between the entrance and exit events. The recording of vehicle colour is a useful way to quickly check against the use of false number plates.

Traffic management

Knowledge of the whereabouts of a vehicle can be most useful. Thus if particular vehicles are detected by a system, police could more rapidly find stolen vehicles. Automatic number plate recognition could be used alongside speed detection systems to automatically create tickets without human input in the image analysis. Similarly, by measuring the average speed of vehicles crossing two points many kilometres apart, drivers that really speed, including those who manage to slow down just prior to the detection point, can be fined.

System considerations

The reason for the previous descriptions was fundamentally to explain that the use of ANPR is very much dictated by the application. It is in this area where Protoclea is offering significant advantages. Having developed a range of ANPR products, Protoclea also has the necessary skills to design or configure solutions to meet the exact needs of the application.

Practical issues

It is worthwhile examining some of the key specifications that have to be considered when evaluating ANPR products:

Cameras

While ordinary CCTV cameras can often be used in these applications they are mostly limited by low resolution. They also require good illumination, perform poorly at night in the presence of headlights, have relatively slow shutter speeds and can therefore only be used for static vehicles.

ANPR cameras generally offer high resolution and added illumination (to reduce headlight glare and the possibility of blinding drivers). This illumination should be sufficient to reflect back from the number plate. They also offer very high shutter speeds (10 000th or 100 000th of a second) for high speed detection and some systems pulse the illuminator to prolong service life and reduce power consumption.

ANPR parameters

The key feature of the OCR part of number plate recognition is the minimum number of pixels required to identify a character. Also critical are the acute angle at which numbers can still be read and the requirement of some systems for numbers to be in a certain format (for example, country specific) to enhance reliability.

The system needs to be able to recognise and ignore the pictures on number plates and be capable of conducting multiple reads (diversity achieves better results - particularly useful in moving applications). Automatic image triggers are required if no other trigger can be provided.

Flexible graphic user interface and system control

Most applications differ in the way they need to use the identified numbers. An ANPR may simply dump the numbers and images into a database for processing by external applications. Others may integrate an application such as security control or access control. The types of applications differ widely, so a system that comes with software support from the supplier is essential to truly customise the application.

Products

Protoclea has launched a number of ANPR systems under the brand SINON. These include:

Low cost systems - 2, 4 or 6 lane systems with fixed software in either access control or security formats.

Flexible systems - also in 2, 4 or 6 lane formats, that offer higher detection quality and can be used for high speed detection. Configurable software is used and special system applications can be developed for users.

Large systems - these can be expanded into very large applications where many entrances and exits service the same protected area. Systems as large as 60 or even 100 lanes can easily be accommodated, keeping data for many months on highly reliable, redundant databases.

Special systems - using high definition cameras, Protoclea is able to offer multilane systems using single cameras.

Share this article:

Further reading: