Thermal Imaging — How it works and what it can be used for

Thermal imaging lets us see something the human eye cannot: the infrared radiation every object emits because of its temperature. A thermal camera turns that invisible radiation into an image, where differences in brightness or colour stand in for differences in how much infrared energy each part of the scene is giving off.

It is a genuinely useful engineering tool — but it is easy to misread. A thermal camera does not simply “see temperature”. Understanding what it actually measures, and where that measurement can mislead you, is what separates a helpful thermal image from a confident wrong conclusion.

What thermal imaging measures

Everything above absolute zero radiates electromagnetic energy. At everyday temperatures, most of that energy sits in the long-wave infrared band — wavelengths of roughly 8–14 micrometres, well beyond what the eye or an ordinary camera can detect. The hotter an object is, the more infrared it radiates, and the more that radiation shifts toward shorter wavelengths.

A thermal camera uses a detector sensitive to this infrared band. Each pixel measures the radiant power reaching it from the corresponding point in the scene. The camera then converts that measured radiance into a temperature and presents it as a greyscale or false-colour image. The colours are a display choice; the underlying quantity is radiated energy.

Apparent temperature vs. true temperature

The temperature a thermal camera reports is best understood as an apparent temperature: the temperature an ideal radiator would need to be to emit the infrared energy the camera measured. Whether that apparent value matches the object’s real surface temperature depends on several things the camera cannot know on its own — most importantly, emissivity, reflections and the path between the object and the lens.

Why a thermal camera does not simply “see temperature”

Three effects stand between “energy measured at the pixel” and “temperature of the surface”.

Emissivity

Emissivity describes how efficiently a surface radiates infrared energy compared with a perfect radiator (a “black body”), on a scale from 0 to 1. Matte, non-metallic materials — painted surfaces, skin, soil, plants, brick — tend to have high emissivity (often around 0.9–0.97) and read fairly truthfully. Bare, shiny metals can have very low emissivity and radiate poorly, so they read far colder than they are unless the camera is told the correct emissivity value.

Two objects at exactly the same true temperature can therefore appear very different in a thermal image simply because their surfaces radiate differently.

Reflections and environmental effects

A low-emissivity surface makes up the difference by reflecting infrared from its surroundings. A shiny panel can show you a mirror-image of a warm object nearby — including the camera operator — rather than its own temperature. Sky, sunlight, heaters and other hot or cold objects in the environment all contribute reflected infrared that the camera happily measures as if it belonged to the target.

The path itself matters too: distance, humidity, and anything in the air between the object and the lens attenuate and add to the signal. Glass and water are effectively opaque to long-wave infrared, so a thermal camera generally cannot see through a window or a body of water — it sees the surface of the glass or water.

Because of all this, quantitative thermography (measuring an actual temperature, not just comparing warm and cold) requires setting emissivity, estimating the reflected apparent temperature, and accounting for distance and conditions. Treated carelessly, a thermal camera produces confident numbers that are simply wrong.

Combining visible and thermal images

Thermal images are rich in temperature information but usually lower in spatial resolution and detail than a normal (visible-light) camera, and they can be hard to interpret on their own — a warm blob is easier to act on when you can also see what it is.

Pairing a thermal camera with a visible camera addresses both problems. The visible image tells you what you are looking at and gives sharp edges and context; the thermal image tells you how warm each part is. Aligning the two — so that a point in the visible image maps to the corresponding point in the thermal image — lets you attach a temperature reading to a specific, recognisable object rather than to an anonymous region of the frame. This alignment is the foundation for many automated uses of thermal imaging.

Representative applications

The same underlying capability — mapping temperature across a scene — supports very different uses across domains:

• Buildings and energy: finding heat loss, missing insulation, air leaks and damp, and inspecting photovoltaic panels for hot cells and faults.
• Electrical and mechanical maintenance: spotting overheating connections, overloaded circuits, worn bearings and failing components before they fail outright (predictive maintenance).
• Process and industrial monitoring: watching temperatures in equipment, material flows or storage where a contact sensor is impractical.
• Safety and security: seeing people, animals or vehicles in darkness, smoke or visual clutter, where a visible camera would see nothing.
• Medical and veterinary screening: highlighting surface temperature patterns as one input among many (never a diagnosis on its own).
• Agriculture and environment: assessing plant and canopy conditions, water stress and surface temperature variation across a field or crop.

These are examples of where the technology fits, not a fixed list — anywhere that temperature differences carry useful information is a candidate.

Combining thermal imaging with detection and edge systems

Thermal imaging becomes far more powerful when it stops being something a person watches and becomes something a system acts on.

• Detection and computer vision can locate and identify objects in the visible image — a component, a person, a piece of fruit — and, through the alignment described above, read the temperature of exactly that object from the thermal image.
• Edge computing lets this run locally, near the cameras, so temperature readings and detections are produced in real time without shipping raw video elsewhere — useful for responsiveness, bandwidth and privacy.
• Data and analytics turn a stream of per-object temperatures into trends, thresholds and alerts, so the system flags what matters instead of producing more images for someone to review.

Combined this way, thermal imaging is less a camera and more a measurement sensor feeding an automated decision.

Limitations and measurement cautions

• It measures surface temperature only — it cannot see inside an object or through glass or water.
• Readings depend on emissivity, reflections, distance and conditions; without accounting for these, apparent temperature can be well off true temperature.
• Shiny and metallic surfaces are especially unreliable without care.
• Thermal detail and resolution are typically coarser than visible imaging.
• It shows temperature, not identity or cause — a warm region tells you where, not why. Interpretation still needs engineering judgement and, often, a second source of information.

Used with these cautions in mind, thermal imaging is a dependable and revealing measurement tool. Used as if the numbers were self-evident, it is a fast route to the wrong answer.

Example explored by Mansonix

Mansonix is exploring a horticultural application that combines visible and thermal imaging. It uses a visible-light camera to detect fruit, including small and tiny tomatoes, and a FLIR thermal camera alongside it.

The approach is to align and overlay the visible and thermal imagery, then use the coordinates of a detection in the visible image to read the corresponding thermal measurement for that same piece of fruit. As explained above, this is a surface thermal measurement — the apparent temperature of the fruit’s outer surface, not an internal fruit temperature. The idea is that pairing “what and where” (from detection) with “how warm at the surface” (from the thermal image) could give growers additional information about crop condition and maturity — the kind of information that might, in principle, support decisions such as nutrient timing or market timing.

This is described as an exploration of the technique. It makes no claim of validated agronomic outcomes, measurement accuracy, commercial deployment, or improvements in yield or revenue; those would require evidence that does not yet exist. What it demonstrates is the engineering pattern this article describes — detection in the visible domain, measurement in the thermal domain, and a deliberate alignment between the two.

Sources & further reading

The technical points in this article — infrared radiation, emissivity, apparent temperature, reflected and environmental effects, and the care needed for accurate measurement — are established thermography fundamentals. For deeper, authoritative treatment:

• Vollmer, M. & Möllmann, K.-P., Infrared Thermal Imaging: Fundamentals, Research and Applications, 2nd ed., Wiley-VCH, 2018 — physics of thermal imaging and its applications.
• Minkina, W. & Dudzik, S., Infrared Thermography: Errors and Uncertainties, Wiley, 2009 — emissivity, reflected temperature and the sources of measurement error.
• Usamentiaga, R. et al., “Infrared Thermography for Temperature Measurement and Non-Destructive Testing,” Sensors, 2014 (open-access review) — measurement principles and application overview.
• Incropera, F. P. et al., Fundamentals of Heat and Mass Transfer, Wiley — black- body radiation, emissivity and the underlying heat-transfer theory.
• ISO 18434-1, Condition monitoring and diagnostics of machines — Thermography — standardised practice for thermographic measurement.