Lead: The payload decides what the flight can report. A 640 sensor with the right field of view will document a failing lug on a distribution structure; a 1280 sensor behind the wrong lens at the wrong altitude will produce a beautiful picture that cannot be measured. This is how to choose a thermal payload from the measurement backwards, and what your airframe will charge you for carrying it.

Start from the measurement, not the megapixel

The number that decides whether your imagery is useful is ground sample distance — how much of the real world falls inside one pixel of the sensor at your working altitude. It comes from two things together: the sensor's pixel count and the lens's field of view.

Two rules follow directly:

  • Narrower field of view puts more pixels on the same target at the same altitude: better GSD, less area per frame.
  • Wider field of view covers more ground per frame and needs a lower altitude — or a bigger sensor — to reach the same GSD.

So the question is never "640 or 1280?" on its own. It is: at what altitude can I legally and safely fly here, and what does that give me per pixel with this lens?

Field of view is a flight-planning decision

  • Narrow (telephoto) lets you stand off — useful near energized equipment, towers, stacks, and anything with a perimeter you would rather stay outside of. Fewer frames, more detail, longer lines to fly.
  • Wide is a coverage lens: roofs, solar arrays, building envelopes, large flat surfaces. More overlap, more frames, faster area rate.
  • Mid is the compromise most inspection programs settle on for a mixed workload.

If your work is mostly one mission type, buy for it. If it is mixed, one mid-FOV payload flown at two altitudes beats two payloads you cannot afford to carry at once.

Radiometric, or a picture of heat?

A radiometric payload records a temperature value for every pixel, which is what lets you report numbers, set isotherms, compare components across a scan, and defend a finding later. A non-radiometric payload produces video and imagery you can interpret visually but cannot measure.

If anyone downstream — a client, an insurer, an engineer — will ask "how hot was it," you need radiometric data and a calibration certificate for the payload. If the deliverable is situational awareness, video is enough.

What the airframe charges you

Thermal payloads are not free to carry:

  • Weight against endurance. Every payload gram comes out of flight time. Check the airframe's published payload capacity against the actual takeoff weight of aircraft, batteries, gimbal, and sensor — then take a conservative slice of the remainder for wind margin.
  • Battery count. A heavier payload usually means more battery pairs to cover the same acreage. Plan the mission in pairs, not minutes.
  • Mount and balance. Some payloads need an adapter or an integration kit; some change the aircraft's center of gravity enough to matter in gusty conditions. Confirm the mount is published as compatible, not "probably fits."

Stabilization and control

A three-axis gimbal holds the sensor steady enough for stills and video at a standoff distance. Beyond that, look for the things automation needs: pitch control that reaches straight down and well past the horizon, a documented way to trigger stills on an interval, and geotagging on every frame so an image can be tied to a location later.

If your program captures hundreds of frames a flight, image geolocation is not a nice-to-have. It is what turns a folder of thermal images into a report.

Match the payload to the mission

  • Solar arrays: radiometric, mid-to-wide field of view, geotagged stills with heavy overlap.
  • Building envelope and roofs: mid field of view, high resolution, oblique passes for wall assemblies.
  • Substations and utility structures: narrow field of view, standoff, radiometric, plus a visible-light channel for context.
  • Search and public safety: wide field of view, high frame rate, and a video feed somebody can watch in real time — resolution matters less than finding the target.

What the manufacturer does not always publish

Not every datasheet lists instantaneous field of view, ground sample distance at a stated altitude, noise-equivalent temperature difference, or radiometric accuracy. When a figure is missing, ask the manufacturer for it — do not infer it from the marketing resolution. A payload with an unpublished IFOV is a payload whose coverage you cannot plan.

Checklist before you buy

  • Does the lens give you the GSD you need at an altitude you can legally and safely fly?
  • Is it radiometric, and does it ship with a calibration certificate?
  • Does the airframe publish this payload as compatible, with the weight and mount confirmed?
  • What is the endurance with that payload fitted, in the wind you actually work in?
  • Are stills geotagged and is interval capture documented?
  • Is the visible-light channel included, or do you need a second payload?

Answer those six and the sensor count becomes the easy part.

Pixels on target: the only number that decides resolvability

Before you compare payloads, work out what will actually land on a pixel at your flying altitude. Ground sample distance for a thermal camera is:

GSD (mm/px) = pixel pitch (µm) × altitude (m) ÷ focal length (mm)

A worked example: a 12 µm pitch detector behind a 25 mm lens, flown at 100 m, gives 12 × 100 ÷ 25 = 48 mm — about 4.8 cm per pixel. At that GSD a 30 cm feature covers roughly six pixels across, which is enough to judge. At 200 m with the same lens it covers three, which is not.

Rules of thumb that follow from it:

  • Five pixels across the smallest feature is the working minimum for a defensible finding. Below that, a hot spot is averaged with its surroundings and reads cooler than it is.
  • Altitude and focal length trade directly. Halve the altitude or double the focal length, and you halve the GSD. Buying a denser detector cannot rescue a lens that is too short for the job.
  • A wider field of view covers ground faster but costs detail. For a fixed sensor size, field of view and detail pull against each other, so pick the field of view from the smallest defect you must see, not from how much ground you want in one picture.
  • "Super resolution" upscaling does not change the native pixel count behind the temperature data. Check the detector, not the output file size.

| Sensor | Typical use | Watch for | | --- | --- | --- | | 320 × 240 class | Wide-area screening, roof and solar overview | Too coarse for small electrical detail at altitude | | 640 × 512 class | General inspection mapping, most commercial work | Needs enough focal length to matter | | 1024 × 768 and above | Small-target detail, long standoff | Cost, weight, and the data volume it generates |