Infrared temperature screening
Thermal imaging cameras are presented as an effective way to detect if someone has one of the symptoms of the coronavirus - a temperature - but it's not that simple.
- Using infrared light to take someone's temperature is a great idea - as long as you don't want it to be particularly precise. Unfortunately, that's exactly what's expected of thermal cameras.
- Thermographic cameras are often coupled with intrusive facial recognition capabilities; some claiming an ability to ID people even whilst wearing masks.
- A person's temperature qualifies as health data under GDPR, requiring data controllers to employ strict safeguards before processing.
- The UK's Information Commissioner's Office stated that employee surveillance must be necessary and proportionate, and that employers should consider less privacy-intrusive means to achieve the same result.
At first glance, infrared temperature checks would appear to provide much-needed reassurance for people concerned about their own health, as well as that of loved ones and colleagues, as the lockdown is lifted. More people are beginning to travel, and are re-entering offices, airports, and other contained public and private spaces. Thermal imaging cameras are presented as an effective way to detect if someone has one of the symptoms of the coronavirus - a temperature.
However, there is little analysis on how this will keep us safe. The aim of this piece is to provide more detail on how thermal cameras use infrared light, and what kind of cameras you might expect to see being deployed. This explainer also highlights why the use of these cameras is problematic.
Let there be (invisible) light!
In 1800, the accomplished German-British scientist Sir William Herschel - on the heels of having discovered the planet Uranus in 1781 - made a discovery which would rock the Scientific World; he discovered invisible light.
Sir William was experimenting with sunlight, testing whether different coloured light had different temperatures, shining through a prism to separate the colours.
Through the use of three thermometers - two outside (one on the left of the colour and one on the right of the colour) to act as "controls", and one used to measure the colour - he showed that each colour of light did, indeed, have a temperature of its own; coolest at blue and warmest at red.
When measuring red light's temperature, he expected the thermometer outside the prism's light to show room temperature -- instead, it measured an even higher temperature than visible red! "Invisible" light - light we can't see with our eyes - had been discovered. This new colour was named infrared ("infra" - below/beneath in Latin).
Whilst we can't see infrared with our eyes, we can easily measure it... as heat. This is why infrared cameras, a relatively low-fi technology, have become the central ingredient of some of the latest technology in the fight against the coronavirus.
Infrared to measure temperature
Using infrared light to take someone's temperature is a great idea - as long as you don't want it to be particularly precise. Unfortunately, that's exactly what's expected of these technologies.
There are three main infrared screening technologies we're seeing deployed around the world:
Spot/laser infrared thermometers
These are usually hand-held devices, which work by reflecting a laser beam off a surface to measure its temperature.
These devices are designed for spot-measuring the temperature of surfaces without touching them - an important design feature as they're for measuring temperatures in the hundreds (sometimes thousands) of degrees.
Beause light beams spread over distance, these thermometers are highly affected by the distance from the surface they're measuring, and require regular calibration against (usually) a reference heat source, from a fixed distance. This calibration should be done right before using it, if you drop it, after you charge it, or after measuring extreme temperatures, like boiling or freezing - again, not uncommon given their second application of measuring food temperatures in kitchens. "Generally, you should calibrate your device every time you feel it’s a bit off."
These are usually hand-held or tripod-mounted cameras which "see" infrared instead of visible light. Originally designed for detecting gas leaks and other hotspots in industrial or mechanical equipment, these cameras use temperature differences to show multicoloured heat maps. The resolution of these cameras is very low (often 640x480) compared to modern visible light cameras, again making them incredibly sensitive to distance from their subject.
There are, however, some thermographic cameras which, when strict procedures are followed, are capable of accurate skin temperature measurement - but it is not as simple as adding a thermal imaging camera to your entrance to filter those with coronavirus.
The camera and its environment must be tightly controlled - temperature, humidity, air currents (fans, air-conditioning, doors, open windows), reflective surfaces (glass/metal), heat sources. The camera must be warmed up for a minimum of 30 minutes before use, and the person being scanned must not have washed/wiped their face or exercised in the 15-30 minutes before being scanned.
Further confounding these cameras is that many are being combined with facial recognition capabilities; some claiming to be able to ID people whilst they're wearing masks, although a recent study suggests this facial recognition is only as accurate as tossing a coin.
This is particularly problematic, considering the invasive nature of facial recognition technology (FRT). FRT relies on the capture, extraction, storage or sharing of people's biometric facial data-often in absence of explicit consent or prior notice. PI is deeply concerned that the use of thermal imaging cameras, coupled with FRT, raises significant problems for our individual freedoms.
Further, while the technology usually boasts a technical accuracy of ~0.5°C, its expected accuracy in real world deployment, the CEO of Movitherm - a thermographic camera company - warns is ±2°C.
For an excellent, in-depth discussion on thermographic cameras, read this NYT piece by their Director of Cinematography.
Crowd/mass scanning cameras
There are no cameras fit for the purpose of crowd scanning. The cameras used are the thermographic cameras discussed above... and as discussed above, they are not fit for this purpose.
As the largest organ with the most surface area, one of skin's many jobs is to help regulate our body's core temperature. If our surroundings are hot, the small blood vessels close to our skin's surface dilate (get bigger), giving them a larger surface area in order to dissipate heat better. If it's really hot, then we sweat, and the evaporation cools us down. Evaporation is endothermic, meaning it takes heat from its surroundings in order to happen - as a result, the surface of your skin is not even remotely uniform in its temperature, and movements of millimetres can swing temperature readings several degrees.
Detect any problems yet?
In addition, facial coverings such as masks (but who wears those anyway, right?), glasses, hats - or even individual hairs on the forehead further degrade accuracy of the scanning - and when we're looking for temperature differences of in fractions of a degree this matters.
Even of those showing a high temperature, and assuming it's not a false positive (these cameras are, after all, only around 90% accurate compared to a rectal thermometer, even after calibration), you've "proved":
- that machine;
- with that calibration;
- in those conditions;
- within its error margin;
- at that precise moment;
- measured that patch of skin to be "feverish".
However, this is subject to the person conducting the scanning being properly trained, as per ISO/TR13154 and the USA's Food and Drug Administration. And of course, they'll promptly follow the standards and confirm the finding using an oral thermometer... I'm sure.
Everything from the common cold to a hangover affects our skin temperature. If someone "passes" the test, is it because they've taken paracetamol recently? Are they over 60, or is it just the time of day? If they "fail" is it from COVID-19, from (some) antidepressants, from HRT or other hormone treatments -- including oral contraception -- or have they just been sitting in the sun on a lovely afternoon?
So what's the problem?
Despite the lack of accuracy and that having a temperature is not necessarily conclusive of a person having coronavirus, the use of infrared temperature scanners and cameras is on the rise.
This technology is already being used at major airports such as London Heathrow, Los Angeles International, Kuwait International and others.
These technologies can infringe individuals’ privacy rights, particularly if the personal information is later shared with airlines, health authorities, or other third parties. We are also concerned that temperature screening is just the first step of more intrusive measures to come, such as health or identity passports or other types of health profiling.
All of these issues raise important questions for fundamental rights and protection of individuals’ personal data, particularly if this scanning is done without prior consent - although even if individuals consent to these scans, it might be problematic to rely on this basis considering the inherent power imbalance between them and the airports/airlines.
The widespread use of thermographic cameras will further normalise surveillance and health and identity profiling, leading to a loss of autonomy and infringement of our privacy rights. Especially since temperature cameras may also be equipped with facial recognition technology, which in this case is intrusive, unnecessary and disproportionate.
Apart from travel companies and airports, numerous employers are also utilising this technology to monitor their worker’s health. For example, Ford, Amazon and the BBC have reportedly introduced thermal scanners. This form of surveillance also raises important issues regarding employees' privacy. Under the GDPR, individuals' temperature would be classified as health data, a 'special category'.
Therefore, simply recording the temperature of employees could amount to processing of special category personal data for the purposes of GDPR, along with the higher safeguards that data protection legislation brings.
any surveillance must be necessary and proportionate and employers should consider less privacy-intrusive means to achieve the same result
The UK's Information Commissioner's Office (ICO) has issued guidance for organisations who may wish to use temperature scanners or any other forms of surveillance on their employees. The ICO guidance emphasises that any surveillance must be necessary and proportionate and that employers should consider less privacy-intrusive means to achieve the same result.
If temperature scanners are imposed as a blanket measure, this could be disproportionate, especially if there are other means of achieving this, for instance by asking an employee how he/she feels on the day. Further, the imbalance of power between an employee and an employer forces the workers to consent to any such measures being implemented. Should they refuse, would they be at risk of being fired? Would they not be allowed to return to the office and lose out on their pay? Considering the unstable economic situation after the lockdown, the workers may not have any choice but to comply with the requests of their employers.
All of these issues, coupled with the lack of accuracy of the temperature scanners, lead to conclusion that these measures are unnecessary, disproportionate and too invasive to be adopted by both governmental agencies and private companies.