What is photogrammetry and how is it used on site

Photogrammetry is the science of producing measurable 3D information from photographs. On UK construction sites, it is the engine behind drone orthomosaics, photogrammetric site models, stockpile volumetrics, and a lot of the imagery you’ll have seen in progress reports. It works because of geometry, processing power, and a careful approach to ground control.

How it works

When two cameras look at the same point in space from different positions, the point appears in different places in each image. The difference between the two — the parallax — depends on how far the point is from the cameras. Triangulating from many overlapping images, software can recover the 3D coordinates of every point that appears in enough of them.

Photogrammetry processing software:

1. Takes a stack of overlapping images.
2. Identifies common features across them.
3. Solves the camera positions and orientations for every shot.
4. Calculates 3D coordinates for every common feature.
5. Builds a dense 3D model and a textured mesh.
6. Generates an orthomosaic — a flattened, geo-referenced image of the captured area.

For survey-grade output, ground control points (clearly marked, surveyed-in targets) are placed in the captured area. The processing software ties the photogrammetric model to the ground control, producing engineering accuracy.

Where photogrammetry beats LiDAR

Photogrammetry uses a camera, which is small, light, and cheap. Drones can carry photogrammetric payloads for sites where a heavier LiDAR sensor would be excessive. Photogrammetric models include true colour by default — the texture of the captured surfaces is the data — which makes them more useful for visual documentation, dispute resolution, and stakeholder communication.

For large external sites with good light and well-textured surfaces, photogrammetry produces a beautiful model fast and cheaply.

Where LiDAR beats photogrammetry

Photogrammetry needs light and surface texture. Dim interiors, plain white walls, glass, water, and shiny surfaces all confound photogrammetric processing. For internal as-built capture and shaded engineering structures, LiDAR is the right choice.

LiDAR also produces points directly, without a processing step that can introduce error in poor conditions. A dense point cloud from a LiDAR scanner is more reliable than a similarly dense cloud from photogrammetry on the same job.

The honest answer is to choose by site condition. External, well-lit, well-textured: photogrammetry. Internal, shaded, plain surfaces: LiDAR. Mixed: both.

What it produces

A photogrammetry deliverable typically includes:

• A geo-referenced orthomosaic of the captured area at high resolution.
• A 3D point cloud of every surface visible to the cameras.
• A textured mesh for visualisation and design use.
• A digital surface model (DSM) showing elevation across the site.
• Volumetrics for stockpiles, earthworks, or other volume calculations.
• The original imagery, geotagged, for inspection use.

Each output ships in standard formats: GeoTIFF for the ortho, LAS or PLY for the cloud, OBJ or FBX for the mesh.

Common construction uses

Site progress capture. A monthly drone flight, processed into an orthomosaic and a short flythrough, gives every stakeholder an honest, time-stamped picture of progress. It costs a fraction of the alternative reporting and resolves more disputes than any volume of written narrative.

Stockpile volumetrics. A drone flight, processed against a known reference surface, produces a defensible stockpile volume in less than a day. It is faster and more accurate than any manual method.

Orthomosaic mapping. A geo-referenced, high-resolution orthomosaic drops directly into GIS or CAD as a base layer for design.

Aerial inspection. A high-resolution photogrammetric capture of a façade, roof, or chimney can be inspected for defects on screen, at any zoom level, without scaffolding or rope access.

Earthworks and quarry work. Bulk material movements, pit and pile geometry, and cut-and-fill calculations all become straightforward with regular photogrammetric capture.

Accuracy in practice

Photogrammetry accuracy depends on three things:

• The image resolution at the surface (governed by sensor, lens, and flight altitude).
• The number and quality of ground control points.
• The processing pipeline.

A typical UK drone photogrammetry job, well controlled, produces accuracy in the range of 2–5 cm in plan and 5–10 cm vertically. For most construction work, that is more than enough. For higher-accuracy work, denser control or LiDAR may be required.

Practical advice

If you are commissioning photogrammetry for the first time, three things matter most:

1. Insist on ground control. A photogrammetric model without control is a pretty picture, not a survey.
2. Match the deliverable to the use. An orthomosaic for a planning application is a different brief from a 3D mesh for design clash detection. Be specific.
3. Look at the surveyor’s previous deliverables. Photogrammetry is highly dependent on the operator’s experience. Past deliverables in similar conditions are the best guide.

The right photogrammetry job on the right site is one of the best-value pieces of survey work in modern construction. The wrong job in the wrong conditions disappoints. Specifying carefully and choosing well makes all the difference.