How to achieve effective camera-based body temperature measurements

With the number of infectious disease outbreaks increasing and the various methods for fever detection seemingly appearing everywhere, if you are considering them it might be worth looking at the International Electrotechnical Commission’s (IEC) guidelines.

IEC / DIN EN 80601-2-59:2017 - guidance for thermography measurements for human febrile temperature screening - includes minimum specifications of resolution, accuracy, as well as the subject placement for the effective measurement of body temperature.

Thermal imaging is one of the most effective methods for non-contact measurement, and there are hundreds of bolometer type detectors out there than can do it. But if they don't measure the right thing, what’s the point? So, before we dive into the guidelines, let’s look into why it’s important to measure the Elevated Body Temperature (EBT) and not the Elevated Skin Temperature (EST).
Skin temperatures can be elevated (or lowered) for many reasons including sweating, being in a hot (or cold) environment, even makeup can play a part! Measuring a location that is as close to the subjects core body temperature as possible, is a much more accurate way of fever screening.

The IEC’s guidelines indicate that the Canthus (inner corner of the eye) is known to provide a reasonable indicative body temperature reading, and that all other methods are inaccurate. According to the international standard “the current evidence indicates that the region medially adjacent to the inner Canthi is the preferred site for fever screening due to the stability of measurement. This is because this region is directly over the inner carotid artery.”

Assuming the optics are very good, to be able to get an accurate measurement from this region 4x4 pixels need to cover it. Any objects smaller than this will appear cooler than they actually are. If the canthus is approximately 3mm in diameter, this would mean a resolution of at least 320x240 would be required to be able to measure the temperature in that region accurately - with the subjects face covering most of the field of view.

The camera needs to be pointing directly at the subjects face, with both canthus regions visible, and there must be no hair or glasses obstructing the face.

Long Wave Infrared cameras can’t “see” through glass, because the transmittance of most glass is very low at the LWIR wavelengths (8–14μm).

This is why measuring large groups of people numbers of people in the cameras field of view is a very inaccurate method of fever screening.

The standards also state that a black body reference must be in focus at the same time as the subjects face, to ensure camera stability (compensating for any potential drift).

It’s important to remember that by following this the system will only be measuring elevated body temperatures in accordance with the guidelines, it’s not detecting any diseases.

How thermal cameras (or, more accurately, thermographic cameras) work

It’s by providing electronic readouts from the Focal Plane Array (FPA), which contains detectors that are heated up by the radiation emitted from the surface of the object.


These emitted rays, which lie in the middle of the infrared range, can only be focused onto the FPA (sensor), by expensive optics made out of zinc salts, germanium, germanium alloys or specially coated mirrors.

camera heat dissipation

Naturally, even with the best cooling mechanisms, the temperature of the Detector, electronics and focussing components will drift during operation. This is caused by heat dissipation that can occur as a result of the elements warming up when they are powered, or simply by the ambient conditions.
This drift affects the accuracy of the temperature reading, and it is one of the main reasons why all thermal imaging cameras have a reading error of approximately ± 2ºC – this increases at temperatures over 100 ºC.

What is blackbody calibration and why would you want it?

calibrated thermal reference

Introducing a calibrated reference point can help mitigate these potential issues by enabling the software to calibrate the scene to the reference. This would improve the accuracy to a more precise value than the camera is capable of.
An object whose radiation emission closely follows Planck’s Law, such that the higher it’s temperature the more radiation it emits, is known as a black body. It’s very non-reflective, and completely absorbs and emits all radiation. A perfect blackbody would have an emissivity value of 1, but no material meets that.
Typically, all blackbody reference emitters work with a heated backplate that has an embedded thermocouple that provides a highly accurate temperature readout. But even with the best engineering, all blackbody calibration sources have a very slightly lower emissivity value than 1.

black body within the camera’s field

The black body needs to be within the same field of the face that is being imaged, with the active area covered with enough pixels to generate accurate results. By having this black body within the camera’s field of view the system can be calibrated in real-time.
With the software constantly comparing the temperature reference to what the camera’s reading is, and making adjustments as necessary. So if the reference is set to 45.0°C and the camera is reading 45.3°C, it would know that the camera’s reading is 0.3°C high.
This constant recalibration dramatically increases the accuracy of the system to ±0.3°C.