2
$\begingroup$

I recently started working with thermal cameras and I was surprised to learn that they don't give you the temperature value of the object you are looking at for each pixel. So I did some reading. I read all the topics on it I could find here. Then some papers I found online. I talked to FLIR. Then I started looking into how they work. Now I think I understand that the microbolometer gives you a value for the temperature of each element of its pixel array. In other words, how much the object it is looking at has heated up each pixel. That's why you can see relative temperature easily.

But then I don't fully understand how you go from that to taking accurate temperature measurements of the object you are looking at. I'm guessing it has to do with understanding the emissivity of the object you are looking at, the temperature of the room, and temperature of the sensor. And then some sort of calibration procedure involving sources (blackbodies?) at known temperatures. I did read through a paper on calibrating, but I left with about the same understanding as when I started.

Anyway I'm really curious how some systems claim to be able to read the temperature of objects using a thermal camera with less than 1C of accuracy. I thought the answer was a camera with a radiometric function, but the expensive ones I look at only measure in one spot! And with +/- 20C of accuracy which isn't much help.

$\endgroup$
7
  • $\begingroup$ "some systems claim..." - Can you post an example of a system that claims this, so we can maybe figure out what the difference is? $\endgroup$ Mar 5, 2020 at 14:12
  • $\begingroup$ @probably_someone oh sure sorry. I was thinking of two things. One I have seen that FLIR cameras like the A20M have "sensitivity" to temperature variations as low as 0.12C. Then I was thinking about this virus/fever scanning system someone showed me. thermoteknix.com/products/oem-thermal-imaging/… $\endgroup$
    – confused
    Mar 5, 2020 at 14:26
  • $\begingroup$ They don't publicly list their temperature range, but it is supposed to be tight enough to discriminate a few degrees for someone with or without fever. (If it really works that is :) ) $\endgroup$
    – confused
    Mar 5, 2020 at 14:27
  • $\begingroup$ Or even like this FLIR handheld camera. It's measuring temperature all over it's image area. docplayer.net/… $\endgroup$
    – confused
    Mar 5, 2020 at 14:30
  • $\begingroup$ This one claims accuracy of 0.1C and full radiometric data. infraredcameras.com/wp-content/uploads/datasheets/… $\endgroup$
    – confused
    Mar 5, 2020 at 15:17

2 Answers 2

1
$\begingroup$

The amount of energy that falls on a pixel does not depend on distance. Suppose at a distance of 1, the parcel of area on the object that the pixel sees is 1. When the distance doubles to 2, the amount of energy that comes to the lens from that area goes down by 4, but the parcel of area on the object that contribute to that pixel goes up by 4. Thus, the amount of energy that falls on the microbolometer pixel is independent of distance. This means that an absolute temperature measurement is related to a power measurement of each pixel, and that can be calibrated.

$\endgroup$
1
$\begingroup$

But then I don't fully understand how you go from that to taking accurate temperature measurements of the object you are looking at. I'm guessing it has to do with understanding the emissivity of the object you are looking at, the temperature of the room, and temperature of the sensor. And then some sort of calibration procedure involving sources (blackbodies?) at known temperatures. I did read through a paper on calibrating, but I left with about the same understanding as when I started.

Exactly. No thermal imaging camera can tell you the temperature of what it is looking at without additional assumptions. Specifically, you need to know the emissivity. From that you can find a blackbody temperature that gives you the emission you measured. To add to the fun, emissivity can be a function of angle, wavelength, and probably others.

Temperature of the room and sensor shouldnt matter for a well designed camera, as they can be calibrated out, but they might degrade the signal to noise ratio. Calibration can often be as simple as measuring a dark scene and a scene of known radiance and interpolating all the intermediate values. This relies on linear sensors, but the kinds of sensors in thermal cameras typically have very linear responses.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.