What Photogrammetry Produces — and What You Do With It
Photogrammetry turns overlapping photographs into measurable 3D data. The processing itself is a solved, well-tooled step — but the files that drop out at the end are where a survey either becomes useful or sits unused on a drive. This guide covers what a photogrammetry run produces, how those outputs differ from LiDAR, what sets their accuracy, and what the rest of the job does with them.
The run produces data, not answers
A photogrammetry run matches features across overlapping photographs, works out where the camera was for each one, and from that reconstructs the position of millions of points on the ground. What it hands back is a set of datasets, each good at a different thing.
Knowing which output answers which question saves a lot of time later — most of the confusion on a site comes from measuring against the wrong one.
| Output | What it is | What it is used for |
|---|---|---|
| Point cloud | Millions of 3D points matched between overlapping photos, usually coloured from the images themselves | Measuring, separating ground from objects, and building a surface |
| DSM | A gridded surface through the top of everything the camera saw: ground, piles, buildings, vegetation | Volumes of piles and structures, line of sight, and the base for an orthomosaic |
| DTM | The bare-earth surface, once objects and vegetation are removed from the DSM | Earthworks, cut and fill, grading, drainage and contours |
| Orthomosaic | The photos corrected for terrain and camera tilt, so every pixel sits at its true map position | Digitising features, plans, and giving the surface a readable backdrop |
| Mesh or textured model | A triangulated skin over the point cloud, textured from the photos | Presentation and inspection more than measurement |
Photogrammetry and LiDAR give you the same kind of thing, differently
Both end in a point cloud and a surface, so downstream they are largely interchangeable. They differ in how they get there, and that difference decides which one suits a site.
Photogrammetry reads the scene from images, so it needs texture and light: a uniform surface — fresh snow, still water, a flat sand face — gives the matcher nothing to lock onto, and it cannot see ground hidden under a canopy, because the camera cannot. In exchange it delivers true colour and an orthomosaic for free, from equipment that is comparatively cheap.
LiDAR ranges to the surface with its own pulses, so it works at night, does not care about texture, and — because a pulse can reach the ground through gaps in foliage and return separately — can recover bare earth under vegetation. It costs more and, on its own, carries no colour.
Accuracy is mostly decided before the flight
By the time you have the outputs, most of the accuracy is already fixed. Ground sample distance sets the finest detail the imagery can resolve; overlap decides how many photos see each point, and thin overlap shows up as noise or holes; ground control ties the whole block to a real coordinate system, and without it a survey can be internally consistent yet sit metres from where it belongs.
This is worth knowing when two surveys of the same site disagree: check the control and the coordinate system before you question the volumes.
The job starts where the processing ends
Producing the outputs is one step; everything a site is actually paid for happens afterwards. Volumes against a design or an earlier survey, cross-sections and quantity tables, contours, slope and drainage analysis, change between two dates, drawings for CAD — all of it operates on the surface and the point cloud, not on the original photographs.
That work is also indifferent to how the data was captured. A DSM from photogrammetry and a DSM from LiDAR are analysed the same way, which is why the capture method and the analysis tool are two separate decisions.
Frequently asked questions
What does a photogrammetry run actually produce?
A point cloud of matched 3D points, a DSM through the top of everything the camera saw, usually a DTM once objects and vegetation are removed, an orthomosaic in which every pixel sits at its true map position, and often a textured mesh. Each answers a different question, so the first step is picking the right one.
What is the difference between photogrammetry and LiDAR?
Both end in a point cloud and a surface. Photogrammetry reconstructs them from overlapping photographs, so it needs texture and light and cannot see ground under a canopy, but it delivers true colour and an orthomosaic from cheaper equipment. LiDAR ranges with its own pulses, so it works at night regardless of texture and can recover bare earth through gaps in foliage, at higher cost and with no colour of its own.
Why is a DSM not the same as a DTM?
A DSM is the surface through the top of everything the camera saw — ground plus buildings, vegetation and equipment. A DTM is the bare-earth surface after those objects are removed. Earthworks, cut and fill and contours should follow the ground, so they need the DTM; volumes of piles and structures are measured on the DSM.
What decides the accuracy of a photogrammetry survey?
Mostly choices made before the flight: ground sample distance sets the finest detail the imagery can resolve, overlap decides how many photos see each point and thin overlap appears as noise or holes, and ground control ties the block to a real coordinate system. Without control a survey can be internally consistent and still sit metres from where it belongs.
Sources
Work with photogrammetry outputs in STREAM
Native Windows · fully offline · free beta