Amogh N P
✦ In loving memory of Amogh N P — Architect · Designer · Visionary ✦
A heritage conservator photographing an ornately carved temple doorway from many overlapping angles with a handheld camera, capturing it for a 3D photogrammetric model, warm daylight.
Unit IVAdvanced Practice & Technique in Conservation

Photogrammetry

3D from photographs — overlap, control and outputs.

≈ 50 min + studio task

Photogrammetry turns ordinary photographs into a measured 3D model. Learn the principle — a point seen in two or more overlapping photos is fixed by parallax and triangulation; classical stereo and modern Structure-from-Motion; the need for scale, control and calibration; the parameters — image overlap and Ground Sample Distance; aerial versus close-range; and the outputs — point cloud, mesh, orthophoto and DEM/DTM/DSM. Try the photogrammetry-overlap explorer.

Learning objectives

By the end of this lesson, you will be able to — mapped to the course outcomes for Advanced Practice & Technique in Conservation:

1
CO4 · Understand

Explain the photogrammetric principle of parallax and triangulation.

2
CO4 · Understand

Distinguish stereo-photogrammetry from Structure-from-Motion.

3
CO4 · Apply

Plan overlap, control and GSD for a heritage capture.

4
CO4 · Analyse

Identify the outputs — point cloud, orthophoto, DEM/DTM/DSM.

Parallax, triangulation, control

The principle & SfM

A point in two or more overlapping photos is fixed by triangulation, exactly as two eyes give depth; modern Structure-from-Motion builds 3D from many ordinary photos — provided there is scale and control.[1, 2]

Parallax & triangulation the 3D point camera 1 camera 2 two viewpoints (parallax) A point seen in two+ overlapping photos from different positions is fixed by intersecting the rays. One photo gives no depth — exactly why you have two eyes. A single image can't make a 3D model.
DiagramPhotogrammetry fixes a 3D point by triangulating the rays from two or more camera positions — the same parallax that gives two eyes depth

Parallax & triangulation

PHOTOGRAMMETRY derives 3D measurement from 2D photographs. The principle: a point visible in TWO OR MORE overlapping photographs taken from DIFFERENT positions can be fixed in 3D by TRIANGULATION — intersecting the rays from each camera to the point. This exploits PARALLAX, the same geometry as human binocular vision: two eyes (two viewpoints) give depth. MISCONCEPTION→correct: 'one photo can give a 3D model' — a single image has no inherent depth or scale; you need at least two overlapping images from different positions.[1, 2]

Interactive

Set the overlap

Slide the forward overlap and watch the photo footprints and the reliability verdict change — heritage close-range work commonly uses about 60–80% forward overlap.

Photogrammetry overlap · set the forward overlap

reliable — every point appears in several photos

Heritage close-range work commonly uses about 60–80% forward overlap (and 30–60% side). More overlap = more redundancy.

the subject (a façade)overlapping photo footprints

A point must appear in two+ overlapping photos to be fixed in 3D — too little overlap leaves holes.

Structure-from-Motion: overlap the subject (a façade) photo 1 photo 2 photo 3 overlap Heavy overlap (≈60–80% forward) means every point appears in several photos — more reliable 3D. SfM auto-matches features across ordinary photos — but needs scale, control & calibration. 'Photogrammetry needs metric cameras' is a myth — ordinary overlapping photos work.
DiagramStructure-from-Motion needs heavily overlapping photos so every point appears in several images
Overlap, GSD, point cloud to orthophoto

Parameters & outputs

Overlap and Ground Sample Distance set the quality; aerial and close-range photogrammetry produce point clouds, meshes, measurable orthophotos and elevation surfaces (DEM/DTM/DSM).[1, 3]

From photos to outputs point clouddense x,y,z meshtextured 3D orthophotomeasurable DEMelevation DSM — surface (+ trees, roofs) DTM — bare earth An ORTHOPHOTO is corrected so it has uniform scale — you can measure it like a map. Measured drawings are then traced from the model and orthophoto. 'An orthophoto is just an aerial photo' is a myth — a raw photo has tilt and relief distortion.
DiagramPhotogrammetry outputs — a point cloud, a textured mesh, a measurable orthophoto, and elevation surfaces DSM, DTM and DEM

Overlap & GSD

Two parameters decide quality. IMAGE OVERLAP: photos must overlap heavily so every point appears in several images — commonly about 60–80% forward (and 30–60% side); heritage close-range work often uses 80%+. More overlap means more redundancy and reliability. GROUND SAMPLE DISTANCE (GSD) is the real-world size that ONE PIXEL covers — a smaller GSD means finer detail; it depends on the sensor, focal length and distance. MISCONCEPTION→correct: 'more resolution is always better' — match GSD and overlap to the required level of detail, not to the maximum the kit allows.[3]

Photogrammetry

At a glance

AspectDetailNote
ViewpointsOne photo: no 3DTwo+ overlapping: 3D by parallax
MethodStereo: oriented pairsSfM: many photos, auto-matched
CameraSfM: ordinary photos work…with overlap, scale & calibration
PlatformAerial/drone: terrain, DEM, orthoClose-range: façades, objects
OrthophotoCorrected, uniform scaleMeasurable, unlike a raw photo
Vocabulary

Key terms

Photogrammetry

Deriving reliable 3D measurement from 2D photographs.

Parallax / triangulation

Depth from two viewpoints; intersecting rays fix a 3D point.

Structure-from-Motion (SfM)

3D from many overlapping ordinary photos, auto-matched by software.

Ground control point / scale bar

Known references that tie a model to real-world scale and coordinates.

Overlap / GSD

How much consecutive photos share / the real size one pixel covers.

Orthophoto / DEM·DTM·DSM

A measurable corrected image / elevation surfaces (terrain, bare earth, surface).

Apply it

Studio task

Plan a photogrammetric capture of a carved temple doorway. State how many photos and what overlap you would use (use the explorer), how you would give the model scale and control, and what outputs you would generate. Then explain in two sentences why a single photo cannot produce a 3D model, and how an orthophoto differs from an ordinary photo.

Check your understanding

Self-assessment

1. Photogrammetry obtains 3D from photographs by —

2. Structure-from-Motion (SfM) is notable because it —

3. An orthophoto differs from a normal aerial photo because it is —

In a nutshell

Recap

Photogrammetry derives 3D from 2D photos — a point in two+ overlapping images is fixed by parallax and triangulation.
Structure-from-Motion builds 3D from many ordinary overlapping photos; metric cameras are not required.
Tie the model to reality with ground control points and scale bars, and calibrate the camera.
Plan overlap (≈60–80% forward) and GSD (the size of one pixel) to the level of detail needed, not the maximum.
Outputs: point cloud, mesh, measurable orthophoto and DEM/DTM/DSM — measured drawings are traced from these.
The evidence

References & further reading

  1. [1]Historic England, Photogrammetric Applications for Cultural Heritage — principle, aerial vs close-range, outputs.
  2. [2]Luhmann et al., Close-Range Photogrammetry and 3D Imaging — triangulation, SfM, control and calibration.
  3. [3]Agisoft / Pix4D documentation — overlap, Ground Sample Distance, orthophoto, DEM/DTM/DSM.
  4. [4]MacDonald (ed.), Digital Heritage: Applying Digital Imaging to Cultural Heritage — heritage photogrammetry in practice.

Further reading

  • Luhmann, Robson, Kyle & Boehm — Close-Range Photogrammetry and 3D Imaging.
  • MacDonald — Digital Heritage: Applying Digital Imaging to Cultural Heritage.
  • R. Agor — Advanced Surveying.

Sources gathered and fact-checked June 2026. Published values vary by source, sample and method — treat as indicative and confirm against the cited standard before structural use.