Skip to content
AkerminMeasure. Monitor. Improve. Compare Products
Learning Center

Emissivity, and why a thermal camera reads polished metal far too cold

A polished metal surface emits only a few percent of the radiation a perfect emitter would, and reflects the rest of what the camera sees from the room. That single fact explains the whole family of absurd readings on bare metal. This article derives it from the two radiance terms and runs the arithmetic out to a worked example.

By the Akermin Editorial Team · 8 min read

As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.

A thermal camera reads polished metal far too cold because polished metal barely emits. Emissivity is the fraction of the radiation a perfect emitter would produce at the same temperature, and for a polished metal surface that fraction is often between 0.03 and 0.10, which means more than ninety percent of what the camera collects from that surface is reflected radiation from the cooler room behind the operator. The camera cannot tell emitted radiation from reflected radiation; it converts the total into a temperature. A copper busbar genuinely running near 77 degrees Celsius in a 22 degree room can therefore display something close to 25 degrees, and the displayed number is not a fault in the camera, it is the honest answer to a question the camera was not asked.

What emissivity is a ratio of

Every surface above absolute zero radiates. A theoretical perfect emitter, a blackbody, radiates the maximum possible amount for its temperature, and the Stefan Boltzmann law gives that amount as a constant multiplied by the fourth power of absolute temperature. Emissivity is the ratio between what a real surface actually radiates and what that perfect emitter would radiate at the same temperature. It runs from zero to one and has no units, because it is a ratio of two powers.

Two consequences follow immediately and both matter in the field. First, for an opaque surface, whatever is not emitted is reflected: reflectivity equals one minus emissivity. A surface with an emissivity of 0.05 is a 95 percent mirror in the infrared, regardless of how it looks to the eye. Polished aluminium looks like metal in visible light and behaves like a mirror in the long wave infrared; matt black paint looks black to the eye and is also nearly black in the infrared, which is a coincidence of names rather than a rule.

Second, emissivity is a property of the surface, not the material. The same copper bar has an emissivity near 0.03 when freshly polished and somewhere between 0.4 and 0.8 once oxidised. Nothing inside the metal changed. That is why a thermal reading of a bare conductor depends on its history, and why two apparently identical bars in the same panel can read differently while sitting at the same temperature.

The complete background to the quantity is in the explanation of emissivity in thermal imaging, and the field level consequences are collected in the note on why shiny metal can mislead a thermal camera.

The two radiance terms a camera cannot separate

What arrives at the lens from a point on an opaque surface is a sum of two things. The first is radiation the surface emitted, which is the emissivity multiplied by what a blackbody at the surface temperature would produce. The second is radiation from everything else in the environment that struck the surface and bounced towards the camera, which is one minus the emissivity multiplied by what a blackbody at the effective temperature of the surroundings would produce.

Written out, the collected radiation equals emissivity times the blackbody radiance at the object temperature, plus one minus emissivity times the blackbody radiance at the surrounding temperature. A third term exists where the air path is long or smoky, since the atmosphere both absorbs and emits, but over a few metres indoors it is negligible.

The detector produces one signal. It cannot attribute part of that signal to emission and part to reflection, so the firmware must be told the emissivity in order to divide the two. Tell it the wrong value and it divides them wrongly, which is the entire mechanism. On a surface with an emissivity of 0.95, getting the split slightly wrong barely matters, because the reflected term is only five percent of the total. On a surface at 0.05, the reflected term is nineteen times the emitted term and the split is everything.

One honest caveat about the arithmetic that follows. An uncooled detector of this class responds over a band, conventionally around 8 to 14 micrometres, rather than across the whole spectrum, and none of the listings in this topic states its spectral band. Integrating the Stefan Boltzmann law over all wavelengths therefore illustrates the mechanism and the rough size of the effect; it does not predict exactly what a particular camera will display.

Working out what a hot busbar reads at five emissivity values

Take a specific situation. A busbar is genuinely at 350 kelvin, which is 76.9 degrees Celsius. It sits in a switchroom whose walls, enclosure and operator are all effectively at 295 kelvin, about 22 degrees Celsius. A camera is told to treat the surface as a perfect emitter, so it converts the total collected radiation directly into a temperature. The table shows what it displays as the real surface emissivity falls.

True surface emissivity Share of the signal that is reflected room Temperature the camera displays Error against the true 76.9 degrees
1.00 None 76.9 degrees Celsius None
0.90 10 percent About 72.4 degrees Celsius About 4.5 degrees low
0.50 50 percent About 52.8 degrees Celsius About 24 degrees low
0.20 80 percent About 35.4 degrees Celsius About 41 degrees low
0.05 95 percent About 25.4 degrees Celsius About 51 degrees low

Read the bottom row again. A bar hot enough to damage insulation displays three and a half degrees above room temperature. There is no alarm threshold that catches that, because the number is not merely inaccurate, it is in the wrong part of the scale entirely. The same geometry makes the top rows reassuring: at 0.90, which covers most painted, plastic and masonry surfaces, leaving the camera on a blackbody assumption costs about four and a half degrees on a target this hot, which is often tolerable for screening.

The curve is also steeply non linear. Going from 1.00 to 0.90 costs four degrees. Going from 0.20 to 0.05 costs another ten. That shape is why the risk concentrates entirely on bare and polished metal, and why nobody worries much about whether plaster is 0.91 or 0.95.

Why the error grows with the gap to the surroundings

The size of the error depends on how far apart the object and its surroundings are, not on the object temperature alone. This follows directly from the two terms. If the object and the surroundings are at the same temperature, the emitted and reflected contributions describe the same temperature, they sum to exactly the blackbody radiance at that temperature, and the displayed value is correct no matter what emissivity is assumed. The error is zero.

Push them apart and the error opens up in proportion to the difference, weighted by one minus the emissivity. Three practical rules come out of that.

  • A low emissivity surface at room temperature reads almost correctly. This is why a shiny panel that has not been energised looks fine and the same panel under load reads absurdly.
  • The reading moves towards the surroundings, not towards zero. A cold pipe in a warm room reads too warm, by the same mechanism and the same arithmetic.
  • Reflections of specific objects matter, not just the average room. A low emissivity surface is a mirror, so a cold window, a bright luminaire or the operator can appear as a localised hot or cold spot on it. Moving two steps sideways and watching whether the spot moves is the test.

This is also the clean way to separate an emissivity problem from a sensitivity problem. A noise floor failure hides small differences in grain; the complete treatment is in the NETD explanation. An emissivity failure produces a confident, stable, wrong number, and a better detector reproduces it more precisely. The two defects look nothing alike once you know what to look for.

Which listings here say the setting exists

All of which is only actionable if the camera lets you enter an emissivity value, and listings in this category are inconsistent about saying whether they do.

Listing Emissivity control in the bullets Other relevant figures published
Thermal Master Thor001 Adjustable emissivity, stated explicitly 35 mK sensitivity, 256 by 192 detector upscaled to 512 by 384, minus 4 to 1022 degrees Fahrenheit, plus or minus 2.7 degrees Fahrenheit, nine point measurement
MiLESEEY TR20 Not stated in the listing bullets Below 40 mK, 320 by 240 detector, 25 Hz, 50 degree lens, plus or minus 2 percent
FOXWELL RT280 Not stated in the listing bullets 120 by 90 detector reconstructed to 240 by 180, 50 by 38 degree field of view, 25 Hz, plus or minus 3.6 degrees Fahrenheit or 2 percent
DCTOPT Not stated in the listing bullets Below 50 mK, 25 Hz, minus 4 to 1022 degrees Fahrenheit, 2 degrees Celsius or 2 percent

Only the Thermal Master Thor001 names the control, in the same bullet as its stated plus or minus 2.7 degrees Fahrenheit accuracy and its nine point measurement function. The MiLESEEY TR20 and the FOXWELL RT280 both publish detailed figures for sensitivity, optics and tolerance and say nothing about emissivity either way. That silence is not evidence the setting is absent. It is evidence that the manufacturer did not consider it a selling point, which is a different thing, and the manual is where the answer lives. How to use the control once you have found it is set out in setting emissivity on a thermal camera.

Turning the physics into a habit on site

None of this requires carrying the arithmetic in your head. It reduces to a short routine.

  • Identify the surface before you read the number. Bare or polished metal means the displayed temperature is unreliable until something is done about it. Paint, plastic, rubber, masonry and timber mean it is approximately usable.
  • Give the camera a high emissivity target instead of a low one. Measure the insulation, the boot, the cable sheath or a painted part of the same assembly. These sit near 0.9 and are usually in good thermal contact with the conductor.
  • Compare rather than read. Where three identical conductors have identical finishes, the hottest of the three is the anomaly even if all three absolute numbers are wrong. Relative comparison is the one technique that survives an unknown emissivity.
  • Move, and watch the spot. If an apparent hot or cold patch on shiny metal tracks your position, it is a reflection.
  • Record what you assumed. A thermal reading without a note of the emissivity setting and the ambient temperature cannot be compared with next year survey of the same panel.

The deeper point is that a thermal camera does not measure temperature. It measures radiation and infers temperature from assumptions you supply. Where those assumptions are close, the inference is good. Where the surface is a mirror, the inference collapses, and the instrument that gets you out of trouble is usually a contact probe or a single point infrared device with a known target, a trade covered in thermal camera against infrared thermometer.

Frequently Asked Questions

Does a more sensitive camera fix an emissivity error?

No, and this is the most useful thing to understand about the two specifications. A low noise floor makes small differences visible. An emissivity error is a systematic offset in the conversion from radiation to temperature, so a more sensitive camera renders the wrong number more crisply. The two specifications address unrelated failure modes.

Why does the error push the reading down rather than up?

Because the object is hotter than its surroundings, which is the usual case for anything you are inspecting. The camera collects emitted radiation from the object plus reflected radiation from the cooler room, and on a low emissivity surface the reflected term dominates. Reverse the situation, with an object colder than its surroundings, and the same mechanism pushes the reading up instead.

Can I set emissivity to 1.00 and subtract a correction afterwards?

You can correct afterwards, but it needs two inputs, not one. The arithmetic requires the true emissivity and the temperature of the surroundings being reflected. Without the second value the correction is indeterminate, because the same measured radiation is consistent with many combinations of object temperature and reflected temperature.

Does the viewing angle change emissivity?

Yes, and it matters most on exactly the surfaces that are already difficult. Emissivity is roughly constant for non metals out to around 45 degrees from perpendicular and falls away beyond that. Polished metals behave less predictably across angle. The practical rule is to view the target as close to square on as the access allows, and to treat a reading taken at a grazing angle as indicative only.

Is a cold reading on a shiny surface always an emissivity error?

Not always, which is why the mechanism is worth knowing rather than memorising a rule. A shiny surface can genuinely be cool, and a reading pulled down by reflection looks the same on screen. The way to separate them is to change the geometry: move the camera and watch whether the apparent temperature of the patch moves with your position. A reflection follows the viewpoint, an emitted temperature does not.