ANPR is a beneficial technology which is too often overlooked for commercial and industrial sites. Often considered the exclusive preserve of law enforcement, the technology actually adds value to many sites with vehicular access control.
When considering ANPR, most people think about law enforcement agencies – the police, the Border Agency, Home Office, etc. – using a system linked to the DVLA database to identify wanted individuals, suspicious vehicles and other known miscreants. Of course, this is a role that automatic numberplate recognition is ideal for, and it has been proven to be very effective in such a role.
It could be argued that the success of ANPR, when utilised in this manner, has led to a somewhat blinkered consideration of its potential for other applications. The technology can be deployed in commercial and industrial applications for a variety of purposes, including security, safety and site management.
In theory, ANPR is akin to access control, and is often used to automate gates and barriers. Approaching vehicles have numberplate details checked against a watch-list. Where a vehicle is known and access is permitted, gates or barriers can automatically be opened.
Of course, there are a number of variations. Certain vehicles can be directed to relevant areas, deliveries or collections can be scheduled out of hours, known visitors can be granted access whilst their host is notified of their arrival, or vehicle details can simply be archived to create an access log.
The benefits are impressive, so long as the system performance is credible.
Image capture
Image capture for ANPR systems is a pivotal part of ensuring that system performance is acceptable. The first step is to consider the ramifications of incorrect or failed capture. If these are too great, then it might be prudent to reconsider the options.
ANPR relies upon capturing the numberplate – typically the front facing one – of a vehicle. Whilst UK numberplates are standardised with regards to font size and shape, it is not unusual for variations to exist. Also, whilst most vehicles will have the numberplate located centrally on the lower part of the front of the vehicle, there will be some exceptions.
ANPR systems use character recognition algorithms, using captured video data to identify and ‘read’ a numberplate. It is therefore important that every effort is made to capture clean, sharp and contrasting images. Whilst many camera manufacturers emphasis the processing they employ is to deliver high quality images, with ANPR cameras you should seek out a device with as little processing as possible.
Everyday video processing – compression, automatic gain control, digital noise reduction, image enhancement – can all make video images appear to be better quality. The important word is ‘appear’! Because the ANPR algorithm is analysing the data, such effects can actually detract from the overall reliability. Therefore, you will require a camera which is capable of delivering high quality images whilst using the minimum level of video processing.
It is also unwise to use cameras with interlace capture. This can either create blur with faster moving vehicles, due to changes in the two interlaced fields, or the ANPR system must rely upon a lower resolution single field image. Cameras with progressive scan chipsets should therefore be the choice.
Shutter speed is another important element when considering ANPR. Many typical CCTV cameras use an average shutter speed of 1/50 second. At such a setting, a camera might be unable to capture suitable images of any vehicle travelling at speeds in excess of a few miles per hour. Often, ANPR will be used at points where traffic slows or temporarily halts. However, this does not mean that shutter speed can be ignored. The philosophy of expecting the unexpected holds true.
It is recommended by Dacolian, an ANPR supplier, that minimum shutter speeds of 1/600 for 30mph traffic, 1/750 for 40mph traffic, 1/900 for 50mph traffic and 1/1300 for 70mph traffic are used. Obviously, the figures need to be rounded upwards to meet the camera settings. Note that the settings should be fixed; automatic shutter speed with a high maximum figure cannot be guaranteed to deliver the right results.
Obviously, as shutter speed increases, so illumination levels need to increase too! The preferred option is to use infrared illumination (along with an IR corrected lens to prevent focus shift). Infrared lighting works well with retroreflective plates, and many dedicated ANPR cameras also include light filters or pulsed illumination to eradicate or compensate for headlight glare which can obscure numberplates. Also, infrared lighting will not dazzle drivers!
There is a fine line when it comes to defining what an ANPR camera is and isn’t! There are some units which have been specifically designed to address many of the issues associated with numberplate recognition. However, there are also a growing number of conventional cameras which have been packaged as ‘ANPR’ units, often by simply including IR illumination. There is nothing wrong with this approach – or indeed with using conventional cameras for ANPR capture – but a number of caveats exist with regard to the specifications of the cameras, and how they can be applied.
A general rule of thumb is that where processing or settings are automatic, often the parameters for self-adjustment will be created with typical wide area surveillance in mind, rather than specialised use such as ANPR. Beware!
The right angles
Once you have a camera capable to delivering the right level of performance, the next task is to ensure that it is mounted correctly. Also, the right lens must be specified. Typically, ANPR system recommendations are that the numberplate should fill 20 per cent of the image width at full resolution.
Cameras should be positioned with a vertical angle of no more than 30 degrees, and a horizontal angle of up to 20 degrees, although this should allow for deviations in vehicle approach. The plate should preferably be in the centre of the screen, and its angle should be such that if a horizontal line was drawn it would pass through both sides of the numberplate.
Most UK numberplates are retroreflective, which means light is reflected directly back towards the source.
The final part
The final part of the ANPR solution is an appropriate software package. Mirasys recommends making several considerations about the choice of software.
If you have a preferred VMS, it probably has an ANPR plug-in. Solicit feedback from colleagues and peers. Asking for reference site details is not always easy – manufacturers often cannot disclose that information.
Consider the supplier’s support. Will they assist with the installation? Do they offer training? How detailed is the documentation? Also what is the total cost of installation, maintenance and ownership? Expect there to be a few days (and nights) of configuration and testing. ANPR has to be set up to meet the site’s requirements.
Configuring ANPR for reliability does take time. Remember that no matter how good the software itself, poor image capture can only result in errors!