A radiometric payload assigns a temperature to every pixel — the difference between “that looks warm” and “that connection is 47 °C above its neighbour.” Getting trustworthy numbers from the air means understanding gain modes, what an accuracy spec actually promises, and the assumptions buried inside the calculation.
Key takeaways
- Radiometric capability is a hardware line item. A non-radiometric payload renders relative contrast only and cannot be made to measure by software.
- Two gain modes cover different ranges: high gain for roughly −20 to 150 °C, low gain out to 550 °C. The wrong mode clips the reading silently.
- Accuracy specs such as ±2 °C or ±2% are laboratory figures at set distance, controlled ambient and known emissivity. Airborne conditions add error the spec does not cover.
- Trust deltas between like components under like load. Absolute airborne values deserve a tolerance of several degrees.
On this page
Gain modes: two ranges, one sensor
Measurement payloads switch the detector between two calibrated ranges. High gain gives fine temperature resolution over roughly −20 to 150 °C, which covers electrical, building and most industrial inspection. Low gain trades sensitivity for headroom up to 550 °C, which is what fire, flare and furnace scenes require.
| Gain mode | Typical range | Use it for | What goes wrong in the other mode |
|---|---|---|---|
| High gain | ≈ −20 to 150 °C | Electrical connections, building envelope, solar, livestock | A flare tip or fire simply saturates — the reading is clipped, not wrong-looking |
| Low gain | up to ≈ 550 °C | Active fire, flares, furnaces, hot process equipment | Small deltas disappear into quantisation; a 3 °C rise is unresolvable |
| Switched mid-flight | — | Scenes containing both | Data before and after the switch is not directly comparable — log the switch |
Clipping is silent. A saturated reading does not announce itself. The image still looks plausible and the number still appears. The only defence is choosing gain for the expected range before launch and re-checking when the scene changes — it is a standing item on the pre-flight checklist.
What accuracy specs actually promise
A specification like ±2 °C or ±2% of reading, whichever is greater, is a laboratory number. It is defined at a set distance, in controlled ambient, against a target of known emissivity, with the sensor thermally stable. Every one of those conditions is different in the air.
| Error source | Direction | Rough field magnitude | Can you control it? |
|---|---|---|---|
| Emissivity assumption | Usually reads low on shiny surfaces | Several °C to tens of °C | Yes — set it, or avoid the surface |
| Atmospheric path | Reads low with distance and humidity | Grows with range | Partly — fly closer |
| Viewing angle | Reads low off-normal | Increases beyond ≈ 45° | Yes — plan the geometry |
| Reflected background | Either direction | Large near hot or cold sources | Partly — change angle or time |
| Sensor stability | Drift until warm | Reduces after warm-up and FFC | Yes — let it stabilise |
The single biggest field error source is emissivity, and it is the one operators most often leave at a default. A polished metal connector with an emissivity of 0.1 will read dramatically cooler than it is, and the same surface will happily show you the temperature of the sky reflected in it. The emissivity guide covers the material values and the workarounds; thermal reflections covers what happens when you ignore them.
The measurement workflow
- Let the payload stabilise thermally and allow an FFC before the first measurement.
- Set emissivity for the dominant surface you are measuring, and note it — the number belongs in the report.
- Choose gain for the expected range, not for the scene you happen to be looking at now.
- Record radiometric files, not screenshots, so analysis can be redone with corrected parameters.
- Capture a reference: an identical component under identical load, or ambient in frame.
- Log distance, angle, ambient temperature and time. Without them the reading is not reproducible.
Recording radiometric files rather than rendered images is the step that most often gets skipped and most often costs a re-flight. A radiometric file carries per-pixel temperature independent of palette, gain display and enhancement, so an emissivity assumption can be corrected six months later at a desk. The handling detail lives in thermal drone data management, and what belongs in the output is set out in report deliverables.
Automating the flag: isotherms and alarms
Isotherm modes highlight every pixel inside a temperature band, and temperature alarms trigger when a threshold is crossed. Both turn a subjective scan into a repeatable one, and both are only as good as the emissivity and gain settings underneath them.
Used well, they let one operator cover a large asset without staring at the feed continuously — the aircraft flags, the operator adjudicates. Used badly, they generate a stream of false positives from reflective surfaces and sun-loaded roofs. Set the threshold against a measured reference on the day, not against a number carried over from the last site. How isotherm alarms work covers the configuration.
Specifying radiometric capability
Radiometric measurement is a hardware capability with a calibration behind it. It cannot be added later in software, and on several payloads it is an explicitly optional thermometry variant rather than a firmware toggle — the OP-125A and LX-9B both list temperature measurement as an option, which means it has to be ordered.
If the report carries numbers, check this line item first. A payload that produces beautiful thermal imagery and no temperature data will pass an unstructured demo and fail the first real inspection. Radiometric versus non-radiometric is the decision to make before comparing anything else, and it belongs in the requirement document — see writing requirements vendors cannot game.
Where measurement is central to the work, budget for annual verification as well. Drift is silent and gradual, and a certificate is part of a defensible deliverable — covered in annual verification.
Related reading
- Emissivity in Drone Thermography: Why Shiny Metal Lies to Your Camera
- Radiometric vs Non-Radiometric Thermal: Which Do You Need?
- Isotherm Alarms and Automated Inspection
- From SD Card to Client Report: Managing Thermal Drone Data
- Annual Verification: When a Thermal Payload Needs Recalibration
- The Thermal Image Processing Pipeline: NUC, FFC, AGC and Detail Enhancement
- Shutters, Blackbodies and Two-Point Correction: The Thermal Calibration Chain
- Substation Drone Inspection: Temperature Rise Criteria in Practice
FAQ
How accurate is drone thermal measurement really?
Within specification under controlled conditions. In the field, treat absolute values as accurate to within several degrees and rely on comparisons between identical components under identical load. The dominant error source is the emissivity assumption, not the sensor.
Can non-radiometric payloads measure temperature?
No. They render relative contrast only, and no software setting changes that. If reports need numbers, radiometric capability is the line item to check before anything else — on some payloads it is an ordered option, not a firmware feature.
What happens if I use the wrong gain mode?
High gain saturates on a hot source, so a flare or fire clips silently and the reading is meaningless. Low gain resolves large ranges but loses the small deltas that electrical and building inspection depend on. Choose for the expected range before launch.
Questions about the technology? Talk to our engineers — we reply within 2 business days.