How accurate is a drone survey, really?

Ask three drone survey companies about accuracy and you will get three confident numbers, none of which mean quite the same thing. Here is how to read them.

Relative versus absolute accuracy

Relative accuracy is how well the model measures distances within itself. Absolute accuracy is how well it sits in the real world, on your project datum. RTK alone gives excellent relative accuracy and can still be a metre out absolutely if the base station coordinate is wrong. These are entirely different failure modes and they should never be quoted as one number.

The number that matters is the check point residual

Ground control points constrain the photogrammetric block. Points used in the adjustment will always fit well - the software optimised for them. That is why check points, deliberately withheld from processing, are the only honest measure. If an accuracy statement does not name how many check points were held out and what their residuals were, it is not a measurement, it is a marketing claim.

What actually degrades accuracy

• Too few control points, or all clustered in one part of the site
• Rolling shutter smearing geometry at survey speed
• Insufficient overlap over featureless ground - water, fresh tarmac, sand
• Flying at an altitude that cannot deliver the required GSD
• Vegetation misclassified as ground, which silently corrupts every downstream quantity

What to ask for

Ask for the check point residuals in all three axes, the number and distribution of ground control, the coordinate system and geoid model, and the reprojection error from the block adjustment. Any competent provider has all four to hand.