A finished orthomosaic is easy to trust for the wrong reason. It looks seamless, the site reads clean corner to corner, and a clean image feels like a correct one. But a map can be visually flawless and still sit a foot off in the wrong spot, because the software that stitches the photos together is built to make the result look continuous, not to guarantee that every pixel lands where it actually belongs on the ground. Before a pad location, a cut-and-fill volume, or a staking decision depends on a map, the question worth asking is not whether it looks right. It’s how its accuracy was actually established and checked.
That question got a more rigorous answer in 2024. The American Society for Photogrammetry and Remote Sensing released an updated edition of its Positional Accuracy Standards for Digital Geospatial Data, replacing a framework that dated back to 1947 and, for the first time, writing dedicated guidance for drone-based mapping rather than folding it into rules built for older aircraft photogrammetry. The update standardized Root Mean Square Error as the primary accuracy metric, tightened the relationship between ground control accuracy and target accuracy, and raised the minimum number of independent checkpoints required to validate a dataset. For anyone verifying a construction site against a design plan, that last change matters most: the standard now expects more proof, not just a cleaner-looking output.
Here’s what a site planning verification workflow looks like when it’s built around that expectation, rather than around how the map appears on a screen.
Size the ground control to the site, not to habit
Ground control points are physical markers, surveyed to known coordinates, that give the photogrammetry solution fixed reference points to anchor against. Without them, a map’s absolute position is only as good as the drone’s onboard GPS, which was never built for construction-grade positioning. How many points a site needs scales with its size and terrain complexity:
- Under about 2 acres: as few as 5 well-distributed markers can be enough.
- 2 to 10 acres: typically 8 to 12 points, spread evenly across the site.
- Larger or more complex sites: proportionally more, with extra points added at terrain breaks and elevation changes.
Distribution matters more than the raw count — a handful of well-placed points on flat ground can outperform a larger number that are clustered together.
Fly to the overlap the site actually requires
Standard sites generally call for roughly 75 to 80 percent frontal overlap and 60 to 70 percent side overlap between images. Tighter urban sites, tall structures, or anywhere the geometry gets complicated call for more — sometimes a double-grid flight pattern to make sure every surface is captured from enough angles to reconstruct cleanly. Altitude and overlap settings aren’t a formality; they’re what determines whether the software has enough information to resolve the site’s actual geometry instead of approximating it.
Validate with checkpoints the model never saw
This is the step the 2024 standard leans on hardest, and it’s the one a lot of verification workflows skip. Some of the surveyed markers on a site should be held back as independent checkpoints — points used to test the finished map’s accuracy rather than to build it. Comparing the map against those held-back points is the only real test of whether the stated accuracy holds up across the whole site, not just near the control points that were used to correct it. A map that hasn’t been checked this way is a map you’re taking on faith.
Overlay the design and read the differences
With a validated map in hand, the actual verification step is straightforward: bring the design drawings in and lay them directly over the current imagery. A pad that drifted from its staked location, a utility alignment that’s off from the plan, an earthwork cut that came in short of grade — these show up as visible offsets against the design rather than as surprises discovered later, during a formal survey or a change-order dispute.
Measure and document what you find
Because a properly controlled orthomosaic is scaled and geometrically corrected, you can pull real measurements directly from it: the distance a curb line shifted, the area still short of grade, the footprint of a staging change. Paired with the date of capture, that becomes a record you can hand to a subcontractor, attach to a pay application, or reference the next time someone asks what the site actually looked like on a given day.
What this doesn’t replace
Orthomosaic verification is a fast, repeatable check between the moments that call for a licensed survey — it is not a substitute for one. Where a boundary, an elevation certification, or another legally required survey deliverable is on the line, that work still belongs to a licensed surveyor. What a well-controlled verification flight gives a project team is a way to catch a drifting condition in days instead of finding out at the next formal milestone.
How Dragonfly Aerials builds this in
Dragonfly Aerials is FAA Part 107 certified, based in Arvada, and serves construction teams across the Colorado Front Range. We size ground control to the site rather than defaulting to one setting for every job, and we validate results against independent checkpoints before a map goes out the door. That process is built to deliver sub-centimeter-level work ready for engineering review, with finished orthomosaics and supporting files returned on a 5–7 business day turnaround.
If your team is relying on aerial mapping to confirm a site matches the plan, our Orthomosaic Mapping & 3D Modeling page covers what’s included, our How It Works page walks through the process end to end, and you can request a quote scoped to your site’s size and tolerances.
Sources:
https://anvil.so/post/photogrammetry-accuracy-standards-explained
https://track3d.ai/blog/ground-control-points-a-complete-guide-for-drone-mapping