How accurate is drone mapping, and how would you know?

Three levels, three different jobs

Drone360 quotes name one of three positioning approaches, because they cost different amounts and answer different questions. A quote that does not name one is not telling you what you are buying.

Relative

The capture is internally consistent. Distances, areas and volumes measured within the site are reliable, but the whole model may sit some distance from its true position on the earth. Right for stockpile volumes, progress comparison and measuring within a single site. Cheapest of the three.

Absolute, using live network corrections

Corrections are streamed to the aircraft in flight from an Alberta base station network, so image positions are recorded in real world coordinates. The result lines up with your design files, your GIS and other datasets. Right for engineering work, corridors and anything that has to integrate with existing data.

Verified against checkpoints

Independent checkpoints are flown and the residual error is calculated and reported with the deliverables. You receive a number rather than a claim. Right for disputes, reserves reporting, regulatory submissions and any project where somebody may later ask how the figure was arrived at.

Why detail is not accuracy

Ground sample distance is how much ground one pixel covers. It is a measure of detail. Accuracy is how close a mapped position sits to its true position. They are independent, and conflating them is the most common way a drone quote overstates itself. A map at 2 cm/px that was flown without corrections can look extraordinarily sharp and still be several metres out of position.

What a real number looks like

On a central Alberta aggregate pit, 1,000 aligned frames captured with live network corrections produced a 2.51 cm/px ground sample distance and a reported 0.01 m RMSE. Both figures are in that sentence for a reason. One tells you what you can see. The other tells you where it is.

Questions worth asking any operator

• Which positioning method is included in this price, and what changes if I move up a level?
• Will the residual error be reported, or am I being asked to assume it?
• Is the quoted figure a ground sample distance or a positional accuracy?
• What reference surface is the volume calculated against?
• Can I have the raw geotagged imagery so my own team can reprocess?

Common questions

How accurate is drone mapping?

It depends entirely on the method used, which is why the method belongs in the quote. A capture flown with live network corrections and verified against independent checkpoints has returned 0.01 m RMSE at a 2.51 cm/px ground sample distance on an Alberta aggregate pit. The same aircraft flown without corrections and without checkpoints can be metres out in absolute position while still looking identical on screen.

What is the difference between GSD and accuracy?

Ground sample distance describes how much detail is in the image. Accuracy describes how close a mapped position is to its true position. A very detailed map can be badly positioned, and a coarse map can be positioned precisely. Quotes that advertise a centimetre GSD and say nothing about positioning are describing detail and letting you infer accuracy.

Do I need ground control points if the drone has RTK?

For georeferencing, RTK handles it. Checkpoints still earn their place when you want the residual error measured and reported with the deliverables rather than assumed, which is what a verified capture gives you.

Which accuracy level does my project need?

If you are measuring within a site, comparing this month to last month or calculating a volume, relative accuracy is sufficient and costs the least. If the data has to line up with a registered plan, a design file or another dataset, you need absolute positioning. If the number will be used in a dispute, a reserves report or a regulatory submission, pay for the verified capture and get the error in writing.