If you have ever opened up a completed orthomosaic and found that a blurry seam was running through the middle of your field or that an area of the site just would not stitch together, the culprit is usually one simple setting: overlap. Not camera resolution, not your drone, and not even altitude on it alone. Overlap.
So, actually how many overlapping photos do you need? By that I mean, it depends on what you are building, is a flat 2D orthomosaic?, is a 3D model? or terrain surface? and it also very much depends on the type of ground you're flying over too. We'll review below the overlap numbers that are right for almost any mission in front and side, how these numbers change slightly for agriculture, construction and forested sites, as well as how to turn a percentage into an actual photo number before you head off.
What Is Front Overlap and Side Overlap?
Well, to the point: there are two kinds of overlap that matter, and they measure differently.
Endlap (or front overlap) means how much of an image repeats in the next image along the same flight line. Overlap between repeat images and one line flight lines is called sidelap. You require both as frontal overlap force the images within a single pass gather and lateral overlap binds the passes to one another in more of a continuous block.
So why does this matter in terms of processing? You learn photogrammetry and all your software assembles an orthomosaic by finding points that are common to many images — tie points — in order to run a bundle adjustment that determines the location of the camera for every image. If a single ground point appears in just two images, it provides the software with only a weak, poorly-constrained estimate of its geographic location. A point appearing in five or six images is indeed a far stronger one. This is directly supported by academic research into tie-point quality: more points per ground feature improve bundle adjustment accuracy, but only if they are evenly dispersed across the image and not concentrated in a single area.
And that is the entire reason why overlap criteria exist. The more overlap we have, the more images observe the same ground feature, which in turn, means a few more tie points, resulting in a tighter and reliable model.
What Is the True Required Overlap?
Most general purpose 2D orthomosaic mapping begins with 75% front overlap and roughly 60-65% side overlap. This is the default in most flight-planning apps, which is sufficient for simple site maps, land inspections and general area calculations.
If you're creating a 3D model — be it a mesh, a dense point cloud, or anything else where vertical accuracy is as much of an issue (or equal to the flat map) — increase those values to 80-85% forward overlap and 70-75% side. Since we want to determine height as well as more classic position, 3D reconstruction requires additional viewing angles on each point to triangulate accurately, so in this case the extra overlap is doing work instead of adding redundancy.
You've got more than enough money to use these projectors without worrying about heading under about 60% front and 30% side overlap. The limit for that is below which photogrammetry software can't identify enough common points to cover the field with a continuous, gapless output — so that's not some target you're aiming for; it's really just your floor.
Is Overlap Different By Type of Terrain?
Yes, and this is where many a pilot has lost their way where the site was needing something more than what he/she would fly with "normal" 75/65.
Agriculture and Open Fields
Crops and bare soil are homogeneous — all a patch of wheat looks just like the one nearby patch of wheat to an algorithm that writes patterns matching (hunting for repeats). Because there tends to be hard features identifiable for image registration their software can lock onto reliable tie points, in order to achieve this, normal practice is 80% or higher both ways respectively.
Forests and Dense Vegetation
Same problem, worse. Canopy is homogeneous, repetitively textured, and wind and sun angle-induced instance shadowing adds noise on top of that. Here you push overlap to 85% or higher.
Construction, Mining, and Uneven Terrain
This one isn't about repetition of features — it's elevation. When you fly a mission at a constant altitude which goes over real relief (a stockpile, graded slope, valley), your overlap on the ground shrinks as the terrain rises towards the drone although that was not what you planned for in your flight plan – we were supposed to get 75%!
A tip from DroneDeploy in this area: clean elevation data needs about 8-9 images over each ground point, while a little drop (typically ≤5 cm) is enough to bring you below that without your knowledge until processing. And this is exactly what those terrain-following flight modes are designed to deal with — they keep a constant height relative to the ground itself (not your takeoff point), keeping your overlap uniform across an entire site.
Calculating How Many Photos You Will Actually Get
After that you can calculate approximately how many images per mission your overlaps have selected. It is very simple logic: work out how much you actually get each new photo (less the overlap), then divide your total site area by that number.
In practice: if you are at 200m x 150m footprint @ your altitude and running 80% front overlap and 70% side overlap each new shot is only adding 20% length & 30% width of "new" ground. Take that smaller footprint and divide it into your total site area, then you've got your approximate number of photos. This is automatically calculated by most flight-planning apps as soon as you set your overlap and boundary, but it's worth understanding the math behind how reducing your overlap from 70% to 85% also nearly doubles up your image count – and processing time.
We experienced this on a drone topographic survey for the feasibility of a dam-site, through terraced farmland and cliff faces at an active watercourse in mountain river valley — not exactly where you want to walk-in and set up the total station! One focal mission flew 403 oblique images at 99m AGL, associated with 12 RTK-monitored ground control points, and achieved an overall RMSE of 5.5cm in accuracy. That kind of accuracy isn't a coincidence from nice weather — the planning overlap and GCP placement needed to match the terrain (not the default) was done with careful planning.
What Is the Minimum Overlap for an Orthomosaic?
The honest floor is about 60% front, 30% side overlap — below that, there simply aren't enough common points available for the software itself to produce a continuous output. But "the software will technically run" is a universe different than "you'll get an accurate map."
Some signs of low overlap are easier to spot in a finished orthomosaic:
- Seams between two flight lines that do not blend cleanly
- Blank or empty patches right at the edges of the coverage area
- A phenomenon inaccurately called "doming" where the whole model curves slightly instead of lying flat
These are all downstream side effects of having too few tie points for the bundle adjustment to operate on. As such, if you see any, the solution almost always lies in those overlap settings on that flight and pretty much never something you patch in post.
Why Does the Overlap Drop on Hilly Sites?
The most common mistake that we see with new pilots is flying a constant-altitude mission and thinking the overlap percentage set in the app is what actually hit the ground.
Not on slope or uneven ground however. When the ground rises toward a drone keeping constant altitude over its takeoff point, the true distance of your camera to the ground shrinks — changing both your actual ground sample distance and your actual overlap - almost always in that dirección junker donde you can afford less. The remedy is an terrain-aware flight plan, climbing in actual time to hold a continuous elevation, over the floor surface itself. If your topographic survey site exhibits meaningful relief, this is not optional however: it's a matter of generating a clean DTM and one with holes in precisely those regions you cared about most.
The Main Point
Overlap is not A SINGLE NUMBER that you set and walk away from.
- An initial value of 75% front / 60-65% side to general 2D mapping
- 80-85% front / 70-75% side for simple, non-armoured vegetation
- Higher numbers for more homogenous terrain or dense foliage / topography
The picture count takes it from there — the more overlap, the more photos and processing time, but that's what really gives the bundle adjustment enough tie points to create a survey-grade result and not a map with holes in it.
If you are unsure that your overlap plan matches your site or if you got a dataset that came back showing visible seams, it's worth getting someone to check the image processing before re-flying! Check out the overlap and GCP planning implementation through real terrain on our portfolio, or let us know if your launch has a flight plan that needs a second set of eyes.
1 comment
good work