How are orthophotos and 3D models made with photogrammetry? Overlap, GSD and ground control
One photo is not a measurement; many photos that see the same point can become one.
Field takeaways
- Photogrammetry computes positions from photos that see the same point from different angles, which is why overlap is essential.
- GSD is the size of one pixel on the ground; it is not accuracy, only a hint at its lower limit.
- Ground control points tie the model to coordinates; check points tell you how well it holds.
- An orthophoto is corrected with a terrain model, a true orthophoto with a surface model; buildings show the difference.
- Water, snow, flat fields and ground under leaves are hard to match; planning happens before the flight.
From photo to coordinates: how it works
Photogrammetry software first finds the same detail (a stone, the corner of a line) in several photos and matches it. When at least two, and preferably more, photos see the same point, the camera positions and viewing directions can be computed together with the point's three-dimensional location. This step is called bundle adjustment; the camera's focal length and lens distortion are usually estimated in the same calculation. Dense matching then produces a point cloud of millions of points, and from it a surface model, an orthophoto and a textured 3D model. Every link in the chain rests on the one before: if matching is weak, none of the later products will be good.
Ground sample distance: how many centimetres is one pixel?
Ground sample distance (GSD; YÖA in Turkish) is the size of one pixel on the ground. For a vertical photo it is roughly: GSD = sensor width × flying height ÷ (focal length × image width). For example, with a 13.2 mm wide sensor, an 8.8 mm focal length and an image 5,472 pixels wide, a photo taken from 100 m has a GSD of about 2.7 cm; at 50 m it halves. In Türkiye, the Large-Scale Map and Map Information Production Regulation sets an upper limit on the GSD of aerial photos: 10 cm for maps and orthophotos at 1:1000 and 5 cm at 1:500. A smaller GSD shows more detail, but on its own it is not accuracy.
Overlap: how many photos should see each point?
Forward overlap is the area shared by two consecutive photos on the same flight line; side overlap is the area shared by two neighbouring lines. The same regulation asks for aerial photography with metric cameras to be planned with at least 70% forward and 30% side overlap, and requires 80% forward and 60% side overlap for true orthophotos. In drone work the cameras are smaller and the flights lower, so most projects plan higher ratios; raising the overlap over woodland and flat, uniform surfaces is what saves the matching. Flying a little beyond the area at the ends of each line makes sure the edges are also seen by enough photos.
Ground control and check points
Ground control points (GCPs) are targets whose coordinates are measured in the field with GNSS or a total station and that can be picked out in the photos; they tie the model to the right coordinate system. They are spread over the corners and edges of the block and inside it; points lined up along a single line leave nothing to stop the model rotating around that axis. Check points take no part in the adjustment and only test the result. The regulation requires the root mean square error at check points after block adjustment to stay within 0.75 × GSD horizontally and 1 × GSD vertically. With a 5 cm GSD that means at most about 3.75 cm horizontally, but only when the criterion is actually met. Even with RTK or PPK on the aircraft, independent check points are still needed.
Orthophoto, true orthophoto and 3D model
An orthophoto removes the displacement caused by terrain, lens distortion and camera tilt, so the photo can be measured like a map. Because it does this with a terrain model (the bare ground), tall buildings and bridges appear to lean and roof edges shift from their true position on the ground. A true orthophoto uses a surface model instead and shows objects in their true planimetric position, which is why it needs more overlap. A textured 3D model is the surface model with the photos draped over it: ideal for inspecting façades, slopes and stockpiles visually, but for measurement it still relies on the control points.
Before the flight: light, surfaces and records
What spoils matching most is the photos themselves. The regulation asks for the sun to be more than 30 degrees above the horizon for aerial photography; long shadows and glare cause trouble. On surfaces with no texture, such as water, snow, freshly ploughed flat fields and uniform roofs, the software finds nothing to match. Watch the shutter speed to avoid motion blur. In Türkiye, drone flights are subject to the registration and permit rules of the Directorate General of Civil Aviation (SHGM). Keep the photos, the control point list and the coordinate system and height type (ellipsoidal or orthometric) the points were measured in together; if height types are mixed, the model can shift up or down by tens of metres.
Turn photos into orthophotos and 3D models.
MapLab Survey 3D produces point clouds, orthophotos and textured 3D models from photos; it opens LAS and LAZ point clouds and produces a terrain surface, cut-and-fill volume and contours as DXF. Windows, on request.
Frequently asked questions
Is GSD the same as accuracy?
No. GSD is the size of a pixel on the ground. Accuracy depends on control points, overlap, camera calibration and processing, and is measured with check points.
How many ground control points are needed?
There is no single number; it depends on the size and shape of the area and the flight method. The corners, edges and interior should be covered, and check points that stay out of the adjustment should be set aside. On public projects, the client's technical specification applies.
If the drone has RTK, are control points still needed?
Fewer control points may be needed, but independent check points are still required to test the result.
What does MapLab Survey 3D do in photogrammetry?
It produces point clouds, orthophotos and textured 3D models from photos; it also opens LAS and LAZ point clouds and produces a terrain surface, cut-and-fill volume and contours as DXF. It runs on Windows and is available on request.
Technical references
The field guidance in this article is aligned with the technical and official references below.
Related: Drone land measurement · LAS and LAZ files · Cut-and-fill calculation · MapLab Survey