As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
A thermal imaging camera works by focusing infrared radiation, roughly in the 8 to 14 micrometre band, onto an array of microbolometer detectors. Each detector element absorbs that radiation and warms very slightly. The camera reads that tiny temperature change as a resistance change, converts it to a digital number, and then uses a calibration curve plus the emissivity setting to convert each pixel into an apparent surface temperature. The displayed image is a visualisation of that calculation, not a direct photograph of heat. The numbers that matter are the detector resolution (how many independent measurements you get), the thermal sensitivity or NETD (the smallest temperature difference the detector can resolve above its own noise), and the accuracy specification (how close the calculated temperature is likely to be to the true surface temperature). The MiLESEEY TR20 and the FOXWELL RT280 both publish 320 by 240 detector arrays; the FLIR One Pro publishes a 160 by 120 detector that it upscales to 480 by 360, which is a different approach to the same problem.
The detector is a tiny thermometer, not a camera sensor
A normal camera sensor counts photons in the visible band. A thermal camera detector does something different: it absorbs infrared photons and measures the resulting rise in its own temperature. The most common detector for handheld and smartphone thermal cameras is the microbolometer, a thin membrane of vanadium oxide or amorphous silicon suspended over a silicon substrate. The membrane is thermally isolated, so a small amount of absorbed infrared energy produces a measurable temperature rise, typically in the millikelvin range.
That temperature rise changes the electrical resistance of the membrane. The camera applies a bias current and reads the resistance change as a voltage. Each detector element, called a pixel, is an independent thermometer. A 320 by 240 array contains 76,800 of them. A 160 by 120 array contains 19,200. That count is the detector resolution, and it is the first number that determines how much spatial detail you can resolve.
The detector does not measure temperature directly. It measures radiance, the total infrared power arriving at the pixel. Converting radiance to temperature requires a calibration curve, usually derived from blackbody measurements at the factory, and a correction for the emissivity of the surface. That conversion is why the displayed image is not the measurement. The measurement is the radiance value plus the calibration plus the emissivity you entered.
Why the lens and the band matter more than the pixel count
Infrared lenses are not glass. Glass absorbs the 8 to 14 micrometre band that thermal cameras use, so the lens is usually made of germanium, zinc selenide or a moulded chalcogenide material. Germanium is dense and expensive, which is why large thermal lenses cost more than the detector in some cameras. The lens focuses the infrared radiance onto the detector array, and its focal length and aperture determine the field of view and the instantaneous field of view, the angle subtended by one detector pixel.
The field of view matters because it sets the trade between coverage and detail. A 50 degree lens covers a wide area in one shot, which is useful for scanning a room or a switchboard. A narrower lens gives more pixels per degree, so a small component fills more of the array. The MiLESEEY TR20 publishes a 50 degree wide-angle lens. The FOXWELL RT280 publishes a 50 by 38 degree field of view. Neither listing publishes the instantaneous field of view or the minimum focus distance, so you cannot calculate the smallest resolvable spot from the bullets alone. The arithmetic is straightforward: if a lens has a 50 degree horizontal field of view and the detector is 320 pixels wide, each pixel subtends 50 divided by 320, which is about 0.156 degrees. At a distance of one metre, that pixel covers about 2.7 millimetres. At three metres, about 8.2 millimetres. A fault smaller than that will be averaged with its surroundings.
The spectral band also matters. Most handheld thermal cameras respond from about 8 to 14 micrometres, which is where room-temperature objects emit most of their energy and where the atmosphere is relatively transparent. A camera optimised for that band will not see through glass, because glass is opaque in that band. It will see the surface temperature of the glass, not what is behind it. That is a physical limit, not a product defect.
The arithmetic between radiance and the number on screen
The processor inside the camera performs three jobs. First, it applies a non-uniformity correction, because each detector pixel has a slightly different response. Second, it converts the corrected radiance value into a temperature using the calibration curve. Third, it maps the temperature values onto a colour palette for display. The palette is a visual aid, not a measurement. Two cameras showing the same scene with different palettes can look very different while reporting the same temperatures.
The emissivity setting is the correction that decides whether your number is right. Emissivity is a number between zero and one that describes how efficiently a surface radiates compared with a perfect blackbody. A matte black surface might be 0.95. Shiny metal might be 0.1. If you leave the camera set to 0.95 and point it at shiny metal, the camera will assume the surface is radiating efficiently when it is not, and the reported temperature will be wrong, usually too low. The camera is not broken; the arithmetic is being fed the wrong input. Setting emissivity correctly is the single most important adjustment for accurate surface temperature.
The accuracy specification, usually written as a percentage plus a fixed offset, describes how close the calculated temperature is likely to be to the true surface temperature under calibration conditions. The MiLESEEY TR20 publishes plus or minus 2 percent. The FOXWELL RT280, the FLIR One Pro do not publish an accuracy figure in the listing bullets. That does not mean they are inaccurate; it means the listing does not give you the number. Where a listing is silent, you cannot assume a value.
Resolution, upscaling and the difference between detail and measurement
Detector resolution and displayed resolution are not the same thing. The FLIR One Pro publishes a 160 by 120 detector and upscales the image to 480 by 360 using Vivid IR. The FLIR One publishes an 80 by 60 detector and upscales to 240 by 180. The FOXWELL RT280 publishes TISR technology that enhances the effective resolution from 120 by 90 to 240 by 180. In each case, the upscaled image contains more pixels than the detector, but the extra pixels are interpolated, not measured. They improve the look of the image and can help you see edges, but they do not create new temperature measurements. The number of independent temperature measurements is still the detector count.
This matters when you are trying to find a small fault. If a hot connection on a terminal block is smaller than one detector pixel, the camera will average its radiance with the cooler metal around it. The reported temperature will be lower than the true temperature of the fault. A higher detector resolution reduces that averaging. A narrower lens reduces it further. Upscaling does not reduce it, because the interpolated pixels are derived from the same detector measurements. Comparing thermal camera resolution properly means comparing detector arrays, not display pixels.
| Product | Detector array (stated) | Displayed or enhanced resolution (stated) | Frame rate (stated) | Accuracy (stated) |
|---|---|---|---|---|
| MiLESEEY TR20 | 320 by 240 | Not stated in the listing bullets | 25 Hz | plus or minus 2 percent |
| FOXWELL RT280 | 120 by 90, enhanced to 240 by 180 | 240 by 180 | 25 Hz | Not stated in the listing bullets |
| FLIR One | 80 by 60, upscaled to 240 by 180 | 240 by 180 | Not stated in the listing bullets | Not stated in the listing bullets |
| FLIR One Pro | 160 by 120, upscaled to 480 by 360 | 480 by 360 | Not stated in the listing bullets | Not stated in the listing bullets |
The table shows why the listing bullets alone cannot tell you which camera is most accurate. Only the MiLESEEY TR20 publishes an accuracy figure. The others publish resolution and features. That is not a reason to dismiss them; it is a reason to read the full manual before you rely on the numbers for a safety-critical decision.
NETD, frame rate and the noise floor
NETD stands for noise equivalent temperature difference. It is the smallest temperature difference the detector can resolve above its own electrical noise, usually expressed in millikelvin. A lower NETD means the camera can distinguish smaller temperature differences. The MiLESEEY TR20 publishes NETD less than 40 millikelvin. The other listings do not publish a NETD figure in the bullets.
NETD matters because every detector has noise, and that noise sets a floor below which a temperature difference is indistinguishable from random variation. If a camera has a NETD of 40 millikelvin, a temperature difference of 20 millikelvin is below the noise floor and will not be reliably visible. A difference of 100 millikelvin is well above it. In building inspection, the temperature differences you are looking for, such as a missing insulation patch or an air leak, are often a few degrees, which is thousands of millikelvin. In electronics, a failing component might be only a fraction of a degree warmer than its neighbours. That is where a low NETD helps.
Frame rate matters for a different reason. A 25 Hz camera updates the image 25 times per second. If you are scanning a moving target, or if you are holding the camera by hand, a faster frame rate gives a smoother image and less motion blur. A slow frame rate can make it harder to see a transient thermal event, such as a relay switching. The frame rate does not affect the accuracy of a static measurement, but it affects how easily you can find the spot to measure.
Where the camera stops being the right instrument
A thermal camera measures surface temperature, and only surface temperature. It does not measure air temperature, core body temperature, or the temperature inside a wall. It does not see through glass, water or most plastics. It does not detect gas unless the gas absorbs in the camera’s spectral band and the camera is tuned to that band. A standard 8 to 14 micrometre camera will not detect a methane leak, because methane does not absorb strongly in that band. A combustible gas sniffer is the right instrument for that job, and a thermal camera is not a substitute.
A thermal camera is also not a moisture meter. It can show a cool patch that might be moisture, but it cannot distinguish moisture from a missing insulation patch, a thermal bridge or an air leak. How a thermal camera sees heat explains the mechanisms, but the interpretation still requires a second instrument, usually a moisture meter or a pin probe. The camera narrows the search; it does not confirm the cause.
Finally, a thermal camera is not an alarm. It displays a temperature map and, on some models, can sound an alarm when a pixel exceeds a threshold. The MiLESEEY TR20 publishes a customizable alarm temperature and a sound and light alarm. That is a useful feature for scanning, but it is not a substitute for a fixed monitoring system with a certified alarm output. If you need continuous protection, a thermal camera is the wrong class of instrument.
The camera is a calculator with a lens
A thermal imaging camera works by measuring infrared radiance, converting it to temperature with a calibration curve, and displaying the result as an image. The detector sets the spatial detail, the NETD sets the noise floor, the lens sets the field of view, and the emissivity setting decides whether the number is right. The display is a visualisation, not the measurement. Once you understand that chain, the specification numbers stop being marketing and start being arithmetic you can check. Choosing a smartphone thermal camera is largely a question of deciding which part of that chain you are willing to compromise on.
Frequently Asked Questions
Does a thermal camera see through walls?
No. A thermal camera measures the surface temperature of the wall. It can show a temperature pattern that suggests what is behind the wall, such as a stud, a pipe or a patch of missing insulation, but it does not see through the material. The pattern is caused by heat conducting through the wall, not by radiation passing through it.
Why does my thermal camera show a low temperature on shiny metal?
Because shiny metal has low emissivity. It radiates little infrared energy of its own and instead reflects the radiance of its surroundings. If the camera is set to an emissivity of 0.95, it assumes the surface is radiating efficiently, so it under-reports the temperature. You can correct this by setting the emissivity lower, or by applying a high-emissivity tape or paint to the surface and measuring that instead.
Is a higher pixel count always better?
Not always, but it usually helps for finding small faults. A higher detector resolution gives more independent temperature measurements, which reduces averaging. However, a high-resolution detector with a poor lens or a high NETD may perform worse than a lower-resolution detector with a good lens and low noise. The detector count is one number in a chain, not the whole chain.
Can a thermal camera measure body temperature accurately?
A standard industrial thermal camera is not a clinical thermometer. It measures surface temperature, and skin temperature is influenced by ambient conditions, sweat and clothing. Some cameras have a dedicated body temperature mode with a reference blackbody, but the listing bullets for the MiLESEEY TR20, the FOXWELL RT280, the FLIR 240x180 and the FLIR 480x360 do not publish a clinical accuracy specification. For medical use, use a device that is labelled for that purpose.
What does NETD less than 40 millikelvin mean in practice?
It means the detector noise is equivalent to less than 0.04 degrees Celsius. A temperature difference smaller than that is below the noise floor and will not be reliably visible. A difference of a few tenths of a degree is well above it. In building inspection, where you are looking for differences of several degrees, a NETD of 40 millikelvin is adequate. In electronics, where you might be looking for a fraction of a degree, a lower NETD is better.
Do I need a thermal camera with a PC software package?
It depends on whether you need to produce reports. The FOXWELL RT280 publishes PC software for point, line and area analysis and automatic report generation, but it is not compatible with Mac. The FLIR One and the FLIR One Pro publish in-app inspection guides and MSX image merging, which is useful for documentation on a phone. If you only need to scan and read temperatures on site, you may not need the software at all.