Seeing what the eye cannot

A thermal camera senses emitted infrared radiation, so it detects things with no visual signature at all: an electrical connection heating up before it fails, water intrusion under a roof membrane, a person in total darkness, the edge of a fire through smoke. For a drone this is one of the highest-return payloads available, because it converts a viewing mission into a detection mission — and detection can be automated, reported, and sold as a deliverable.

The four specifications that decide quality

Resolution

Thermal sensor resolution is dramatically lower than visible-light sensor resolution — you are working with a fraction of the pixels. That makes flight altitude critical: a small object simply will not register as enough pixels to be identifiable. Thermal missions fly lower and slower than visible mapping missions, which affects area coverage and therefore your operating cost per acre or per mile.

Thermal sensitivity (NETD)

The smallest temperature difference the sensor can distinguish, usually quoted in millikelvin. This is what decides whether you find a connection running slightly warm or only a spectacular failure that would have been obvious anyway. For inspection work it is often the most important number on the datasheet.

Radiometric versus non-radiometric

A non-radiometric camera gives you a relative image — hotter and colder. A radiometric camera gives calibrated absolute values per pixel, which is what allows measurement, thresholds, and trending between visits. Radiometric costs meaningfully more, and without it you have comparison but not measurement. Decide this based on whether your report contains numbers.

Frame rate — and export control

Higher frame rates improve tracking of moving subjects, and they also trigger regulatory attention. Uncooled thermal imagers above certain frame rate and performance thresholds are controlled for export from the United States under the EAR, with some higher-performance and military-specific sensors falling under ITAR. This directly constrains which sensor you can specify and which countries you can sell into, so resolve it before you design the payload around a part — the same class of market-access question as the compliance marks discussed in product safety testing requirements.

Why thermal measurements mislead

The most common mistake is reading the number on screen as the true temperature. In reality the value is distorted by four things:

  • Emissivity. A polished metal surface and a painted surface at identical temperatures look completely different. Shiny metal is the classic trap in electrical inspection.
  • Reflections from the sun, the sky, or nearby hot objects.
  • Viewing angle. Apparent temperature falls off at oblique angles.
  • Distance and atmosphere — humidity and path length attenuate the signal.

Professional practice therefore follows simple rules: fly when there is no direct solar loading on the target, compare identical items to each other rather than to an absolute threshold, and hunt for anomalies — the one component hotter than its neighbors. Anomaly detection is robust against all four error sources at once, which is why it is the basis of nearly every commercial thermal inspection product.

Fusing thermal with visible imaging

A thermal image alone is hard to interpret; it is genuinely difficult to know what you are looking at. Professional systems therefore carry two cameras and align them, so a hot spot can be overlaid on the visible image and the operator immediately sees that it is that insulator or that window. Good alignment requires geometric calibration between the two optical paths, accurate time synchronization, and stable mechanical registration — which means the two sensors must be rigidly mounted relative to each other on a stabilized gimbal, and the calibration has to survive vibration and thermal expansion. This is where a lot of homemade dual-sensor payloads fall apart.

Four mature applications

  • Electrical asset inspection — transmission and distribution lines, transformers, switchgear, and panels. A loose or corroded connection runs hot, and finding it prevents an outage. This is the largest commercial use.
  • Solar farms — a failed cell, string, or diode heats up, and one aircraft can scan an entire site in hours instead of a crew walking rows for days. A very large and well-established market.
  • Building envelope — thermal bridges, missing or wet insulation, and moisture intrusion, valuable for both energy audits and construction defect documentation.
  • Search and public safety — locating a person or animal in darkness or vegetation, and mapping the active edge of a wildfire through smoke.

From image to report — where the money is

The business value is not the imagery; it is the report: a severity-ranked anomaly list, each item with precise coordinates, a context photo, a measured or relative delta, and a comparison against the previous inspection. Producing that requires automated anomaly detection, projection into geographic coordinates, and change detection between visits — the capability stack described in AI in hardware products. A company that sells inspection reports occupies a far stronger commercial position than one selling flight hours, which is essentially the argument in hardware as a service.

Integration realities on the aircraft

Thermal cameras are among the most expensive payloads you will carry, and they are awkward mechanically. The sensor and its processing electronics generate heat while simultaneously being sensitive to it, so placement is not trivial — you cannot bury it next to the power stage, and you cannot let motor exhaust wash over the lens. Add mass and center-of-gravity effects on the gimbal, vibration isolation, and the electrical interface, and payload integration becomes a real design task rather than a bolt-on. Plan the mass and power budget with the aircraft, as covered in motor and propeller selection, and route the interface through the flight stack deliberately — see choosing a flight controller.

Thermal also sits on the ground looking up in the counter-drone world, and it is a core sensor for agricultural water stress work. Every aircraft you build for sale in the US must also meet Remote ID requirements, and program cost expectations are in drone development cost.

Export control classification is a legal determination. Projects House is an engineering firm, not a law firm — this article is educational only, and you should have qualified export counsel classify your specific sensor and product before selling internationally.

Specify the payload against the report you intend to sell

Projects House designs and integrates drone payloads — sensor selection, gimbal and thermal design, calibration between visible and infrared channels, and the processing chain that turns imagery into a deliverable report. If you are building an inspection product, tell us what the report needs to say through our contact form and we will work backward to the hardware.