How to: light for CCTV and facial recognition
June 2019, by Gemma Snelling, Marketing Manager
June 2019, by Gemma Snelling, Marketing Manager
HD, ultra HD and 4K, are just a few of the latest developments in CCTV, but even with these technological advancements, it still relies on a lot of light to be effective - so when it does go wrong it is more often than not the result of poor or badly designed lighting.
Our top tip - if tasked with a security lighting scheme where close-up, high-quality imagery is required to enable an individual to be recognised and identified2, is to swap your lighting designers cap and start thinking more like a photographer.
We asked our lighting applications team to list their top 5 recommendations for designing for CCTV and facial recognition:
We understand that in an energy-conscious age, we are all trying to reduce the amount of energy (and thus light) we use, but, if you want to be able to clearly identify a criminal on your CCTV, a different mindset must apply. You are going to need to use lots of light.
Lighting for CCTV should be based on the inverse square rule: if you double the distance to the subject being lit, you will need FOUR times the original light. Additional lighting can be used to create an evenly lit scene in the camera’s field of view, as this will ensure captured images are not too dark or washed out. However, this is also true in reverse. It is usually recommended that there be no more than a 3:1 contrast ratio of minimum or maximum illumination within an artificially lit scene.
There are a number of traps a lighting designer can fall into, here are just a few:
Occurs when a camera is faced with anything bright in the background, including the sun and street lighting; causing loss of distinguishable features in images. To eliminate this risk, you need to think carefully about the position of cameras.
Positioning lights should be behind CCTV cameras, not pointing at them.
Flicker and motion blur
There are many different types of CCTV so make sure you’re aware of the challenges associated with each, such as motion blur. Some cameras may say that they operate in low light conditions, but only produce monochrome footage and if you want to capture a car or a person moving, it’s unlikely you’ll get a clear view of their face or the car’s license plate.
Working to the right standards
The police state that over 80% of the video evidence that they collect fails to meet the required standard. To identify a person for evidence (that will stand up in court), the Home Office requires that for traditional PAL-based CCTV systems, a person of average height needs to fill the full height of the screen and show their face. With picture quality and detail sufficient enough to identify the individual without reasonable doubt. The lighting therefore needs to be good enough that an operator can identify two faces from a standard chart of 12 racial groups in 30 seconds.
High-pressure (or low-pressure) sodium lighting is terrible for CCTV with nearly all colour information lost, so if the area in question has this - consider upgrading to LED which is far more effective at producing true lifelike colours. The colour rendering index (CRI) should be over 80. Many CCTV manufacturers will recommend cool colour temperatures as high as 5000K, but this is contrary to public preference – we would not recommend any higher than 4000K - but the right balance must be found for the area and technology in place.
When an emergency arises, we rely on emergency lighting to kick in. This isn’t enough for CCTV to continue effectively; as BS 5266-1: 2016 states light levels will drop to 0.5 lx, which is much too low for CCTV. You will need a secondary light source or use a camera with integral lights to ensure continuous footage during a mains failure.
Our best piece of advice. When it comes to CCTV, vertical illumination is king for creating facial recognition. However, areas where CCTV has been introduced will have been designed to meet horizontal illumination targets, which casts shadows on people, leaving the faces of subjects wearing hats and hoods totally dark. To counter this you need to think like a photographer, we recommend mounting additional luminaires at face height - 1.6m.
More light = more reliable data
Move away from your traditional view on light levels – you need to measure at the subject’s face. Not on horizontal surfaces. Aiming to achieve 30lx from the floor level measured vertically. In practice, this can be tricky, but 20 to 30lx should be achievable. Go a step further and use at least two points of light source to create some facial modelling.
Ren - a bollard, which isn’t a typical solution for lighting for CCTV we appreciate, but Ren offers enhanced vertical illumination - perfect for facial recognition
Sabre - performance floodlight for uniformly lit large open areas, specify with intelligent control options such as a PIR sensor which will detect motion/presence
Kirium Wall & Sephora Wall - CCTV friendly, low flicker premium drivers with integral 3 hours maintained emergency - perfect for building and perimeter lighting - for mounting at 3-6M
Kova - finally for underpass applications or areas prone to vandalism, consider Kova - our IK10+ rated anti-vandal luminaire