Positioning
On this page 9
A detection is only useful if somebody can drive to it. This page is about getting a centimetre-class position onto every frame without buying equipment you do not need.
What accuracy each capability actually needs
| Capability | Needed | Why |
|---|---|---|
| Wildlife search | 2 to 3 m | Somebody walks to a warm blob. Standard GNSS is fine |
| Bird deterrence | 2 to 3 m | |
| Livestock patrol | 2 to 3 m | |
| Disease, pest, fertilisation | 0.3 to 1 m | Zones are tens of metres across |
| Field mapping | 2 to 5 cm | Every later flight is registered against this base |
| Weed control | 2 to 5 cm | A prescription cell is metres across and the sprayer has its own error |
| Yield counting | 2 to 5 cm | Overlap de-duplication depends on knowing where each frame was |
| Seeding | 10 to 30 cm | The spread pattern is wider than the error |
Only half the catalog needs RTK. Do not put a €300 receiver and a correction subscription on a wildlife aircraft.
The three ways to get RTK
1. A network correction service. Start here
Somebody else's base stations, delivered over the internet as NTRIP.
| Service | Coverage | Cost |
|---|---|---|
| SAPOS | Germany, run by the state survey authorities | Varies by federal state. Several states publish it free of charge; check your Landesamt für Vermessung |
| Centipede RTK | France and neighbouring countries, community run | Free |
| RTK2go | Worldwide, community casters of varying quality | Free |
| Commercial networks | Wide, guaranteed | €300 to €1,500 per year |
| Onocoy | Growing, incentivised community network | Low |
For most of Germany and much of western Europe there is a free or cheap correction stream within the 10 to 30 km baseline that RTK wants. This is almost always the right answer.
2. Your own base station
Worth it when the site has no mobile data, when you fly the same site constantly, or when you need a local reference for survey checkpoints.
| Item | EU source | US source | Indicative |
|---|---|---|---|
| Emlid Reach RS3 survey receiver | Emlid | Emlid | €2,500 to €3,300 |
| ArduSimple simpleRTK2B as a fixed base | ArduSimple | SparkFun | €200 to €400 |
| Survey antenna and ground plane | ArduSimple | SparkFun | €100 to €300 |
| Radio link, 868 MHz in the EU | RFDesign | RFDesign | €250 pair |
A DIY base is a few hundred euro and works. The commercial receiver buys you a known antenna phase centre, a tripod that survives a season, and a support number in April.
3. Post-processed kinematic
Log raw observations on the aircraft, download a reference station's observations afterwards, and solve on the ground with RTKLIB (BSD). No live link needed at all.
This is the right answer for restoration sites and anywhere with no coverage. It costs nothing but processing time, and German SAPOS and many European networks publish RINEX observation files for exactly this.
Receivers
| Choice | Notes | EU source | US source | Indicative |
|---|---|---|---|---|
| ArduSimple simpleRTK2B | u-blox ZED-F9P, the standard | ArduSimple | SparkFun | €200 to €300 |
| SparkFun GPS-RTK-SD | Same chip, logs raw for PPK | Mouser EU | SparkFun | €250 |
| CubePilot Here4 | Integrates cleanly with a Cube flight controller | Drone Parts Center | GetFPV | €250 to €350 |
| Emlid Reach M2 | Rover with easy PPK workflow | Emlid | Emlid | €700 to €900 |
The antenna matters as much as the receiver. A multi-band antenna with a proper ground plane, mounted away from the flight controller and the video transmitter, is the difference between a fix and a float solution.
Getting the position onto the frame
This is where accuracy is usually lost, not in the receiver.
| Step | What to do |
|---|---|
| Hardware trigger | Fire the shutter from the flight controller and log the event, or read the camera's hot shoe feedback |
| Interpolate | The exposure happens between GNSS epochs. Interpolate the trajectory to the exposure timestamp |
| Lever arm | Measure the offset from the antenna phase centre to the camera's optical centre and apply it. It is typically 10 to 25 cm, which is five times your RTK error |
| Orientation | The IMU's attitude at exposure, for projecting frame corners onto the ground |
| Verify | Ground checkpoints. Not the same points used to fix the solution |
Skipping the lever arm is the classic error. A 20 cm offset is an order of magnitude larger than the 2 cm the receiver reports, and it is systematic, so it does not average out.
Checkpoints
Survey three to five permanent, identifiable points per site with a survey receiver, once. Every subsequent flight can be checked against them in seconds, and it is the only way to find out that the RTK was lying: an incorrectly entered base coordinate produces a beautifully precise map in the wrong place, and nothing in the reconstruction will tell you.
In-house software
| Stage | What we run | Licence |
|---|---|---|
| NTRIP client | On the ground station or the companion computer | own code, or str2str from RTKLIB |
| PPK solving | RTKLIB or rtklib-explorer | BSD |
| Geotagging | Trigger log plus trajectory, with lever arm and interpolation | own code |
| Coordinate transforms | PROJ, via pyproj or GDAL | MIT |
| Checkpoint reporting | Residuals per flight, stored with the mission | own code |
Storing the checkpoint residual on every Mission row costs nothing and turns
"is the map any good" into a number you can show a customer.