"How many photos do I need?" is the first question every new drone mapper asks, and it is the wrong one. Photo count is not a setting. It is what falls out of three decisions you make first: how fine the result has to be, how much the photos overlap, and how large the site is.
Get those three right and the count takes care of itself. Get them wrong and you can fly six hundred photos of a two hectare field and still end up with holes in the model.
Key takeaways
Ground sample distance is set by flight height and the camera. It decides how fine the output can be.
Overlap decides whether the reconstruction works at all. 75 to 80 percent forward and 65 to 70 percent side is a safe default.
Photo count is a consequence: site area divided by the ground each photo adds after overlap.
More photos never fix bad overlap, and they never make the result finer than the flight height allows.
The photo count is an output, not an input
Photogrammetry works by seeing the same point on the ground from many angles. Every point in the model has to appear in several photos taken from meaningfully different positions. Seen in two photos, its position is weak. Seen in nine, it is well constrained.
That is why overlap matters more than volume. Two hundred photos flown in a tight grid at 80 percent overlap reconstruct cleanly. Six hundred flown in one long line, or with a gap where the battery was swapped, will not.
So the planning order is: pick the detail you need, derive the flight height, pick overlap from the scene, and let the count come out at the end.
Ground sample distance: what flight height actually buys you
Ground sample distance, or GSD, is the size of one output pixel on the ground. At 3 cm/px, one pixel of your orthomosaic covers a 3 by 3 cm square of the site. It is the single number that decides whether a crack, a survey nail or a paint line will be visible in the result.
GSD comes from three things: the sensor's pixel size, the lens focal length, and the flight height above ground.
GSD (cm/px) = sensor width (mm) x flight height (m) x 100 / (focal length (mm) x image width (px))The practical consequence is simple. Halve the flight height and you halve the GSD, but you also quadruple the number of photos needed for the same area. That trade is the real cost of detail, and it is worth pricing before the flight rather than after.
What you need to see | Typical GSD | Consequence |
|---|---|---|
Site overview, stockpile shapes, field boundaries | 4 to 6 cm/px | Fast flights, few photos, large area per battery |
Construction progress, kerbs, drainage, markings | 2 to 3 cm/px | The common default for surveying work |
Cracks, bolts, small defects, as built detail | Under 1 cm/px | Low, slow flights, photo count rises sharply |
Exporting the orthomosaic at a finer resolution than the flight allows does not add detail, it resamples. Flown at 3 cm/px, a 1 cm/px export is three times the file size and exactly the same information.
Overlap: the number that decides whether it reconstructs
Overlap is expressed as two percentages. Forward overlap is how much consecutive photos along a flight line share. Side overlap is how much neighbouring flight lines share. Both matter, and they fail in different ways: too little forward overlap breaks the chain of matched points along the line, while too little side overlap leaves seams and warped strips between lines.
Scene | Forward | Side | Why |
|---|---|---|---|
Open bare ground, quarry, field | 75% | 65% | Plenty of texture, matching is easy |
Construction site, mixed built area | 80% | 70% | Height changes and occlusions need more views |
Dense vegetation, forest, uniform crop | 85% | 75% | Repetitive texture, weak features, moving leaves |
Facades, towers, anything vertical | 85% or more | 80% or more | Oblique passes, every surface needs several angles |
Two habits are worth building. Plan overlap for the worst part of the site, not the average, since one low overlap strip over the area the client cares about ruins an otherwise good flight. And keep the flight height constant relative to the ground on sloped sites. At a fixed altitude above the take off point, the top of a hillside is photographed much closer than the bottom, and the effective overlap changes with it.
Check the flight before you plan the next one
Drop the photos in and the data check reads the EXIF headers in your browser: estimated GSD, forward and side overlap, and a warning if the overlap is too low to reconstruct cleanly. The free plan takes up to 150 photos per project.
Check your photo setPutting it together: a worked example
Take a 20 megapixel camera with a 5472 by 3648 pixel sensor, flown so the resulting GSD is 3 cm/px.
Each photo covers 5472 x 0.03 = 164 metres across the flight line, and 3648 x 0.03 = 109 metres along it.
At 80 percent forward overlap, each new photo advances the aircraft by 109 x 0.20 = 21.8 metres. At 70 percent side overlap, each new flight line sits 164 x 0.30 = 49.2 metres from the last.
So every photo contributes roughly 21.8 x 49.2 = 1072 square metres of new ground. A 10 hectare site is 100,000 square metres, which needs about 93 photos. Add a margin for the turns at the ends of the lines and for a buffer around the area of interest, and 110 to 120 photos is a realistic plan.
Nobody chose "100 photos" there. It came out of a detail requirement, an overlap setting and an area. Change the GSD to 1.5 cm/px and the same site needs roughly four times as many photos, because both the along track and across track footprints halve.
Add a buffer, always
Fly past the edge of the area you care about. Photogrammetry is weakest at the flight boundary, since points near the edge appear in far fewer photos, which shows up as noisy, warped elevation in the corners of the DSM. One extra flight line on each side is cheaper than a second site visit.
What goes wrong, and how to spot it before processing
Most failed reconstructions were already visible in the photo set before anyone pressed start. Four checks catch nearly all of them.
Missing geotags. If the photos carry no GPS coordinates in EXIF, the result cannot be placed on a map, and processing has nothing to anchor to. This is usually a drone setting, or the result of the photos being edited and re exported by a tool that strips EXIF. Always upload the originals.
Only sea level altitude. Some cameras write GPS altitude above mean sea level but not height above the take off point. Without a relative altitude, the ground footprint of each photo is unknown, so GSD and overlap cannot be estimated from the file alone. The flight may still be fine, but you lose the ability to check it in advance.
Low overlap. An estimated forward overlap in the 50s means gaps are likely. Reflying beats spending processing time first.
Motion blur and low light. Photogrammetry needs sharp, well exposed frames. Flying fast in low light produces images that look acceptable at thumbnail size and match badly at full resolution.
What the photo count does not change
A larger photo set does not improve accuracy beyond what the flight height and the georeferencing allow. It does not fix a site flown in bad light, and it does not turn a nadir only flight into a usable facade model. It does make processing slower.
Frequently asked questions
What is the minimum number of photos for a drone map?
There is no single figure, but a handful of photos will not reconstruct. Photogrammetry needs many overlapping views of each point, so even a small subject usually needs tens of frames. If a site is small enough to need under about 30 photos, overlap is the limiting factor, not count: fly a proper grid over it rather than a few frames from one position.
Is 70 percent overlap enough?
For open, well textured ground, 75 percent forward and 65 percent side usually works. 70 percent forward is on the edge and leaves no margin for wind, turns or a slightly off nominal flight height. Over vegetation or anything with repetitive texture, push both numbers up rather than down.
Does flying higher with a better camera give the same result?
For GSD, roughly yes: a larger sensor or longer lens can hold the same ground sample distance from higher up, covering more area per photo. What it does not fix is geometry: flying higher means shallower angles between views of the same point, weakening the height solution, so elevation degrades faster than the image does.
How long does it take to process a few hundred photos?
It depends on the photo count, the requested outputs and the processing quality, so treat any single figure with suspicion. Run one representative project at the settings you plan to use as standard, and use its runtime as your baseline for quoting.
Photo count is the last number to think about, not the first. Decide the detail you need, derive the flight height, set overlap from the scene rather than from habit, and add a buffer around the area of interest. The count that comes out of that arithmetic is the right one, and it will be defensible when the client asks why the flight took as long as it did.