Sensors

The payload is what a flight is. This page covers the four sensor families the catalog uses, at three budgets each.

RGB

The workhorse. Resolution and shutter type matter more than anything else.

What actually matters

Ground resolution

GSD (cm/px) = altitude (m) x sensor pixel pitch (µm) / focal length (mm) / 10

This table is why the cheap sensor is not free. Coverage is swath times speed, and swath is pixels across times ground resolution. At 1 cm/px the Pi camera's swath is 14.6 m against the 24 MP camera's 60 m, so it covers roughly a quarter of the ground per battery and flies a third of the altitude to do it.

Multispectral

Five bands, calibrated, with a downwelling light sensor. See disease detection for the band table and why calibration is not optional.

The DIY route works and is genuinely used in research, but budget for the band registration work: separate bodies mean separate lenses, separate viewpoints and a per-frame alignment step that the integrated cameras do in hardware. It is a week of engineering plus permanent calibration discipline to save several thousand euro. Worth it once, as a way of understanding the pipeline. Rarely worth it as the production payload.

The calibration kit, which nobody budgets for

An uncalibrated multispectral flight is a picture. A calibrated one is a measurement. The whole change detection argument depends on which one you have.

Thermal

Radiometric, always. A palette JPEG has thrown the measurement away.

Resolution buys hectares

Thermal cores share a 12 µm pixel pitch, so with a lens matched to the array, doubling the pixels across doubles both the usable altitude and the swath. Area per battery is swath times speed, so it roughly doubles too.

That table is the whole argument for the 640 sensor on a wildlife search, where the operating window is two hours long.

Export control

Thermal cores above certain resolution and frame rate thresholds are export controlled in both the US and the EU. Buying a 640 x 512 core is routine but involves paperwork and end-use statements. Buying from an EU distributor into the EU removes most of it.

LiDAR

Only two capabilities genuinely need it: canopy height for weed control discrimination, and field mapping where a canopy hides the ground. Photogrammetry cannot see under trees; LiDAR can.

A cheap LiDAR gives you points. Survey-grade LiDAR is an IMU problem, not a laser problem: without a tightly coupled inertial and GNSS trajectory, the point cloud is precise and inaccurate. Budget for the IMU and the post-processing, or buy the height-hold sensor and use photogrammetry for the map.

Choosing for each capability