Surveying with GNSS and drones

Traditional surveying carries coordinates forward from point to point along lines of sight. GNSS removes the line of sight but adds a condition: the sky has to be open. Here is how the methods differ, how a drone gets centimetre positions, and the one check that decides whether any of it can be trusted.

One receiver is not surveying

A single receiver gives you metres. That is nowhere near what surveying asks for.

GNSS surveying gets its accuracy from two or more receivers observing at the same time. A nearby receiver sees the signals through the same ionosphere and the same atmosphere, so it carries the same error. Take the difference and the shared part cancels.

What remains is the relationship between the two points, called the baseline. GNSS surveying is the work of stacking baselines until the coordinates close. Put a known point at one end and the other end gets a coordinate.

Choosing a method

Static: both receivers sit still for an hour or more. Millimetres. The satellite geometry changes while you wait, which is exactly what buys the accuracy. Results come later, not on site.

Rapid static: around twenty minutes, roughly a centimetre.

Kinematic: measured while moving, a few centimetres, for taking many points.

RTK: corrections arrive live and coordinates appear on site in seconds. This is what you want when a stake goes in the ground today.

Network RTK: no base station of your own; a service combines several permanent stations so that accuracy holds even far from any one of them.

What actually limits accuracy

How much sky you can see. Low satellites blocked means fewer usable satellites and lopsided geometry. Height suffers first.

Reflections. Walls, metal roofs, vehicles and water surfaces. A metal plate under the antenna cuts what bounces up from below.

Baseline length. The further apart, the less the two receivers share, so less error cancels. Within 10 km is comfortable; past 20 km it gets hard.

Antenna height. Measure from the mark to the antenna reference point and write it down. One centimetre wrong here is one centimetre wrong in every height afterwards, and no software will ever tell you. Measure it twice.

How a drone gets centimetre positions

Photographs and point clouds carry shape, not location. Something has to tie that shape to the ground.

The old way is targets: lay marked boards across the site and survey each one. They appear in the photographs, and the computation ties shape to coordinates. The cost is time, and on a large site laying and collecting them can take half a day.

The other way is to put an RTK receiver on the aircraft. Every photograph then carries the position of the moment it was taken, and most of the targets can go.

Do not remove all of them. Keep a few as check points; without them there is nothing left to verify against.

The check everyone skips

After the result comes out, measure a known point again and compare. It takes minutes.

Skip it and the first sign of an error arrives at the very end, when the deliverable is already in someone else's hands.

If it does not match, work through it in order: was the fix solution actually fixed, did the corrections keep arriving, was the base too far, was the sky open enough, and was the antenna height recorded correctly. That last one is invisible to every calculation.