Key takeaways
- An orthomosaic is not a photograph. It is hundreds of photos geometrically corrected so that scale is consistent across the image, which is what makes measuring on it possible at all.
- It is safe to measure horizontal distances and areas on the ground surface. It is not safe to measure heights, anything above ground level, or anything hidden under canopy, water or a roof.
- Ground sample distance sets the floor on detail. If one pixel covers a few centimetres of ground, no amount of zooming recovers detail finer than that — it just makes the pixels bigger.
- Learn the four artefacts that fool people: buildings leaning outward from the flight centre, seam lines where photos were joined, smeared thin structures like power lines and railings, and ghosted vehicles that moved between frames.
An orthomosaic is usually the first thing a client opens, and it is the deliverable that gets misread most often — because it looks exactly like a photograph of the site, so people read it like one. It is not a photograph, and the differences are the whole point.
This is what an ortho actually is, what you can measure on it, what you cannot, and the handful of visual artefacts worth being able to recognise before you make a decision on one.
What an orthomosaic actually is
A single aerial photo has perspective. The camera sees the ground directly beneath it straight down and everything else at an increasing angle, so scale changes across the frame and anything with height leans away from the centre. You cannot measure on that image, because a centimetre in one corner is not the same distance on the ground as a centimetre in the middle.
Orthorectification fixes this. The processing software works out where the camera was for every photo, builds a model of the ground surface, and re-projects each image onto that surface so the result is as if every point were viewed from straight above. Scale becomes consistent. Then the corrected images are blended into one continuous image — the mosaic part — and tied to a coordinate system, so every pixel has a real-world position.
The result is a map that happens to look like a picture. Our orthomosaic glossary entry covers the short definition, and the photogrammetry entry covers how the underlying model gets built.
What is safe to measure
On a well-produced orthomosaic, these are reliable:
- Horizontal distances on the ground. Fence line to building corner, road width, the length of a laydown area.
- Areas and outlines on the ground. Disturbed area, gravel coverage, the footprint of a pad, the extent of a spill or a flooded zone.
- Positions. Where something is, in whatever coordinate system the deliverable was produced in.
- Counts and inventory. Vehicles, pipe joints, panels, culverts — anything you can see and identify.
- Change between two dated captures, provided both were flown to the same reference and the comparison is honest about what moved.
These are the measurements that survive because they all live on the ground plane, which is the surface the image was corrected onto.
What is not safe to measure
Everything that leaves the ground plane.
- Heights and elevations. An ortho has no vertical dimension. Height comes from the surface model or the point cloud that was produced alongside it, not from the image.
- Volumes. Same reason. A stockpile volume is computed from a surface against a defined base — see our base surface entry — and never from the ortho, even though the ortho is what you use to draw the pile's outline.
- Anything on a vertical face. A wall, a tank shell, the side of a bin. A straight-down view sees the top of it and almost none of the face.
- Rooftop dimensions, at least not without care. On a standard ortho the roof has been projected onto the ground surface, so its footprint is displaced. More on this below.
- Anything under something else. Ground under tree canopy, the bed under water, anything inside or beneath a structure. The ortho shows the top of what was there on the day.
Ground sample distance is the detail floor
Every ortho has a ground sample distance: the real-world size of one pixel. It is set by the camera and the flight altitude, and it is decided before the aircraft leaves the ground.
GSD is a hard limit on what the image can tell you. If one pixel covers several centimetres of ground, a crack narrower than that does not appear as a crack — it appears, if at all, as a faint tonal difference, or not at all. Zooming in does not recover it. This is the single most common disappointment with an aerial deliverable, and it is entirely predictable at the scoping stage: say what the smallest thing you need to see is, and the flight gets planned to resolve it. Our GSD glossary entry covers the arithmetic.
The corollary is that measuring to a precision finer than a pixel is self-deception, however many decimal places the software reports.
The four artefacts worth recognising
1. Buildings and tall objects lean
This is the one that surprises people most. On a standard orthomosaic, the ground is corrected properly but objects standing above the ground are not — a tall structure appears to lean outward, away from the point the camera was directly over. Building tops sit offset from their bases, and you can often see the side of a tank or a silo that a true straight-down view would not show.
This is normal and expected, not a defect. It matters because a roof outline traced off a standard ortho is not the building's footprint — it is the roof, displaced. A "true ortho" product corrects above-ground objects as well, which costs more processing and needs more imagery; it is worth asking for when building footprints are the deliverable, and unnecessary when they are not.
2. Seam lines
The mosaic is assembled from many photos, and the software chooses where to cut between neighbours. Along those seams you may see a sudden change in brightness, a shadow that jumps, or a linear feature such as a road marking that steps sideways slightly. Seams tend to be worst where the ground had the most height variation, and around the edges of the flown area where there was less overlap to work with.
If you are measuring something that crosses a visible seam, treat that measurement with more caution than one taken well inside a single source photo.
3. Smeared thin structures
Power lines, fences, handrails, guy wires, thin masts and open steel are the classic failure case for photogrammetry. They are too thin to be reconstructed cleanly, so they appear smeared, broken into fragments, or absent entirely — and the ground behind them may be distorted where the software tried to fit a surface to them.
The practical rule: an orthomosaic is not the tool for inspecting thin linear structures. Oblique close-range imagery is, and it is a different flight.
4. Ghosts and moving objects
Anything that moved during the flight is captured at different places in different photos. A truck driving through the site can appear twice, appear half-blended into the ground, or disappear at a seam. Water almost never reconstructs well — moving or reflective surfaces have no stable texture to match — so ponds, dugouts and rivers often show as noisy, distorted or oddly flat patches.
None of these are errors in the survey. They are inherent to how the product is built, and knowing that stops you raising them as defects.
Shadows, light and the fact that it is one moment
An ortho is a record of one flight on one day. Shadows are baked in: they can hide a defect entirely, and they can be mistaken for one — a long shadow off a stack reads convincingly as a dark stain on the ground. Overcast light gives flatter, more even imagery with fewer shadows and is often better for interpretation, even though it looks less attractive.
The same applies to colour. Unless the capture was radiometrically calibrated for a specific purpose, the colours are the camera's rendering, not measured values. Two orthos of the same site flown a month apart will not match in tone, and that difference is lighting and processing rather than a change on the ground.
Two kinds of accuracy, and which one you are relying on
When someone quotes an accuracy figure for an ortho, ask which one they mean.
- Relative accuracy is how well the product measures against itself — the distance between two points inside the image. Usually good, and it is what most on-screen measurements rely on.
- Absolute accuracy is how well the product sits in the real world, in the stated coordinate system. This depends on the positioning method and the control used, and it is what matters when the deliverable has to line up with existing survey, design or GIS data.
A survey can be internally excellent and absolutely wrong. Our posts on how many ground control points you actually need and RTK vs PPK cover how absolute accuracy is achieved and verified.
Check what it is sitting on before you use it
An orthomosaic carries a coordinate reference, and a GeoTIFF should declare it. If the reference is missing or wrong, the image lands in the wrong place, or in the right place while measuring distances that do not agree with your drawings. This is worth two minutes at the start rather than a week of confusion later — our post on which coordinate system your drone survey should be in covers how to specify it, and it belongs in the scope before the flight.
A two-minute sanity check on any new ortho
- Measure something you already know. A building you have dimensions for, a road allowance, a parking bay. If it comes out right, the scale and the reference are working.
- Look at the edges. The boundary of the flown area has the least overlap and the most distortion. If the thing you care about sits near the edge, be cautious — and next time, ask for the flight area to extend past the area of interest.
- Zoom to the smallest feature that matters and ask whether it is genuinely resolved or just a few pixels you are interpreting generously.
- Find one seam. It tells you how many source photos are in play and how well they blended.
- Check the date and the light. Then decide whether the answer you are about to give depends on conditions that have since changed.
What the ortho should arrive with
On its own, an orthomosaic answers "what is there and where". The questions about height, volume and change need the products that come with it: a surface model, contours, a point cloud, or a report built from them. Our post on drone survey deliverables explained covers what each one is for, and how 3D mapping differs from 3D modeling covers the distinction people most often blur.
Read the ortho as a measured map of the ground on a particular day, and it is one of the most useful documents a site can have. Read it as a photograph and it will eventually mislead you.