If you're thinking of investing in a number plate recognition system, perhaps to restrict traffic into your site, reduce losses or analyse the frequency of visits to your premises, there are a number of factors you'll need to consider to get a system that delivers reliable detection and information.
Known as Automatic Number Plate Recognition (ANPR), License Plate Recognition (LPR) or Number Plate Recognition (NPR) is a powerful technology that can help you remotely detect and log vehicles that pass your cameras.
A camera for LPR should be set up so it can capture the small area of the number plate consistently and clearly in all lighting conditions to provide the highest recognition rate possible. This intense focus comes at a CPU cost, as the processing resources needed for this type of analysis can be high, so you'll need to ensure you have enough capacity if you're using a server-based solution, such as with Milestone’s LPR plugin. Some cameras built specifically for LPR do the processing on-board, so this can help in keeping processing costs down.
Choosing the right location
Mount the camera as close as possible to where the number plate will be and use only optical zoom to get the right view. You want the number plate to be perpendicular to the camera or just a few degrees off for the best detection accuracy. The higher the angle away from the perpendicular the lower the likely accuracy.
We recommend having a maximum vertical angle of between 15-20° for the height of the camera and a maximum horizontal angle of 15-25° to the number plate (though always check the specs of the LPR solution you choose). Ideally, place your camera so vehicles progress up or down your view, this gives your software the maximum opportunity to recognise a full number plate or even more than one at a time.
Your camera should cover only the vehicle and road, with your view showing as little of the surroundings as possible. Irrelevant moving objects should be avoided. You should consider the location/environment for best detection accuracy for example:
- Speed of the vehicles (the slower the better – some traffic-calming measures may be needed)
- Number plate size and position within the camera’s field of view (the larger the better, up to a point)
- Lighting conditions (is the area well-lit with either white light or infrared?)
- Camera exposure and shutter speed (optimised for the lighting and vehicle speed and set to avoid over or under-exposing reflective number plates)
- Resolution of the camera (higher resolutions may be more accurate but require higher processing)
It is important to understand what the particular requirements are for the solutions you choose as they are not all the same.
Choosing the right camera is key
There are some locations where there is only one place to put your camera and therefore at different parts of the year, it could be looking straight into the sun resulting in overexposed images. If this sounds like your site, choosing a camera with Wide Dynamic Range (WDR) can compensate for overexposure. Axis has just released a number of Q series cameras which offer forensic WDR and high resolution like Q1645-LE or Q3517-LVE and there are other similar cameras on the market. Surveillance cameras not specifically built for LPR can be just as good at performing this function if chosen correctly.
If you're looking for 24-hour surveillance, then your camera will need to have good low-light visibility and infrared (IR) performance. You could also consider using external illumination such as Raytec Illuminators to improve your image recognition. There are cameras which are designed specifically for LPR. These tend to be a little more expensive than other cameras, yet cameras such as the ones mentioned above can work perfectly well in most scenarios.
How fast is a vehicle going when it passes your camera? If you're trying to capture a car travelling at speed then you'll require a camera with high frame rate capability. High frame rate in combination with multi-megapixel resolution, will help to produce accurate results. For example, using 60 fps rather than 30 fps gives double the frames for your software to analyse.
Your analysis options
The first and most common option is to integrate number plate detection into an existing VMS, such as Milestone XProtect. In this case, a small additional cost adds LPR capability to the system and we find that the detection accuracy when using Milestone, provided the camera is situated correctly, is very high indeed.
A second method is to use Axis cameras as we suggest above, adding in a third-party ACAP (Axis Camera Application Platform), a piece of analytical software running on board the camera, to perform the processing (there are a few options here).
Finally, there are LPR systems available where the camera is connected to an NVR which processes the detections. The downside here is that you often cannot integrate this with a larger system, where you can easily with the first two options.
Your software
The key to the effectiveness of any number plate recognition system is the software doing the detection.
As the global number one for VMS over the last ten years, Milestone has strong number plate recognition through their LPR plugin. The plugin lets users capture, and analyse footage as well as create events and notifications from the interface. For example, a number plate could be compared to a list of approved plates and granted access by activating a gate, or it could be used to analyse the volume of traffic to see how effective a marketing campaign has been. Similarly, blacklisted number plates can easily flag an alarm in the system to be actioned, or to deny access.
Some of the strengths of Milestone are the high accuracy rate, the open-platform choice of cameras, the possibility to hook any part of the system in with LPR, the ability to investigate your data (including partial plates) and to create PDF reports.
The Axis ACAP offers a number of different apps to allow you to use your camera for LPR.
Closing thoughts
With cameras and VMS getting more powerful every year, number plate recognition technology is falling into the price bracket of more businesses. The purpose of this blog is to include the main areas to consider to make a system successful. Correctly specifying and positioning your camera within your scene is important as it will not only give enough time for the software to recognise a plate but also gives you the widest choice of cameras. Finally, the right software can help you accurately monitor vehicles ultimately helping you protect and grow your business.