What is Photogrammetry? How Photos Become Maps and 3D Models
Photogrammetry is the science of taking measurements from photographs. Photograph a site from many positions with enough overlap, and software can work out where every camera was and where every visible point lies in three dimensions. Drone photogrammetry applies this to mapping: a few hundred photos become an orthomosaic, an elevation model, a point cloud and a textured 3D model.
The idea behind it
Hold up a finger and look at it with one eye, then the other: it seems to jump against the background. How far it jumps tells you how far away it is. Photogrammetry does the same with cameras. A point that appears in two photos taken from different positions can be triangulated, and a point that appears in many photos is triangulated many times over, which is what makes the result measurable instead of a guess.
Two things have to be solved together: where each photo was taken from and how its lens bends the picture, and where each point on the ground is. Because neither is known at the start, the software solves for both at once, and the process is called structure from motion followed by bundle adjustment.
How photogrammetry works, step by step
Software finds distinctive spots in each photo, matches the same spots across overlapping photos, and uses the matches to solve the camera positions, the lens and a first sparse set of 3D points. It then estimates a depth for almost every pixel against neighbouring photos and fuses the estimates into a dense point cloud. From the cloud come a surface mesh and an elevation raster, and projecting the photos onto that surface gives the orthomosaic.
| Step | What happens | What comes out | Stream Capture stage |
|---|---|---|---|
| Read photos | Metadata, positions and camera data are read; bad photos are flagged | Camera groups, coverage | Data |
| Match and solve | The same points are found across photos; camera positions and lens are solved | Camera poses, sparse cloud, reprojection error | Align |
| Densify | Depth is estimated for almost every pixel and fused | Dense point cloud | Dense Cloud |
| Build surfaces | The cloud becomes a mesh and an elevation raster | 3D model, DSM, DTM | Mesh + Texture, DSM, DTM |
| Orthorectify | Each photo is projected onto the surface and blended | Orthomosaic | Orthophoto |
What you need to get a good result
Overlap above all: each point on the ground should appear in several photos from different positions. At least three overlapping photos are needed to start, and a real map needs many more. Sharp, well-exposed photos with visible texture matter too, because the software matches texture.
To put the result in a real coordinate system at the right scale you also need positions: from an RTK or PPK drone, from ground control points measured on the site, or from both. Without them the model is internally consistent but floats in space.
What photogrammetry is used for
Topographic maps and elevation models, stockpile and earthwork volumes, change between two flights, site documentation, and 3D models of structures and terrain. Mining and quarrying, construction, land survey, agriculture and archaeology all use it. The outputs are standard files: GeoTIFF rasters, LAS or LAZ point clouds, DXF contours and OBJ or GLB models.
What it cannot do
It sees only the surface the cameras saw, so it cannot see through dense canopy to the ground beneath, and bare-earth models under vegetation are estimates. It struggles on surfaces with no texture, such as still water or fresh snow, and on things that move between photos. And it measures nothing in absolute terms without control: scale and position come from known positions, not from the photos alone.
Frequently asked questions
What is photogrammetry in simple terms?
Measuring from photographs. Overlapping photos taken from different positions are matched, and the software triangulates where each camera was and where each visible point is in 3D.
How does photogrammetry software work?
It matches the same points across overlapping photos, solves the camera positions and lens, estimates depth for almost every pixel and fuses it into a dense point cloud, then builds surfaces, an elevation model and an orthomosaic from the cloud.
What is drone photogrammetry?
Photogrammetry with photos taken from a drone, usually in a grid flight over a site, to produce maps, elevation models, point clouds and 3D models.
How many photos do you need for photogrammetry?
At least three overlapping photos to start; a map of a real site needs far more. Count by coverage: every point should appear in several photos.
Is photogrammetry accurate?
It can be very accurate, but accuracy depends on camera, overlap, positioning and control, and it is measured at independent check points. Height is normally weaker than horizontal position.
What is the difference between photogrammetry and LiDAR?
Photogrammetry derives 3D from photographs and sees only visible surfaces; LiDAR measures with its own laser light and can return through gaps in vegetation. For open ground and structures photogrammetry is often enough.
Turn your own flight into an orthomosaic, DSM and 3D model
Native Windows · fully offline · free beta