Photogrammetry

Photogrammetry is a method that determines the shape, dimensions and position of an object from photographs taken from different directions. A point captured in at least two images is located by intersecting the rays from those images, much as in triangulation. The software simultaneously solves for the position and orientation of the camera in every image, a step known as bundle adjustment.

Metrology (industrial) photogrammetry uses coded and uncoded retroreflective targets stuck onto the object, a calibrated camera and scale bars of known length. Its output is not surfaces but highly accurate coordinates of the targets. Depending on the system, accuracy is on the order of hundredths of a millimeter per meter of object size, so even a part several meters long can be measured to within tenths of a millimeter or better. The same principle is used by optical systems that measure displacement and deformation during load tests.

Image-based photogrammetry (structure from motion), by contrast, turns dozens to hundreds of ordinary photos into a dense point cloud or a polygon mesh with a color texture. It needs no targets, but the surface must have a distinct pattern. Accuracy depends on image resolution, distance and camera network geometry and is considerably lower than that of industrial 3D scanners: typically tenths of a millimeter to millimeters on objects, and centimeters on buildings and terrain captured by drone.

In 3D scanning of large objects, photogrammetry is often combined with a scanner. Reference targets across the whole object are measured photogrammetrically first, and the scanner then places each scan into this fixed framework. This limits the error that builds up as scans are stitched together, which is the main source of inaccuracy on large parts.

Metrology photogrammetry pays off on large objects where the overall dimensions and the relative position of points matter: welded frames, body panels, molds, fixtures, steel structures and large machine assemblies. It is used to check the position of clamping and mounting points, to measure deformation and as a reference framework for 3D scanning parts several meters in size.

Image-based photogrammetry suits the documentation of buildings, factory halls, terrain, sculptures and historical artifacts, where completeness and color texture matter more than hundredths of a millimeter. It needs little equipment, but taking and processing the images requires experience and time.

Without a scale reference, a photogrammetric model does not have the correct size. Measurement requires calibrated scale bars, usually at least two placed in different directions. Quick models from smartphone apps have uncertain accuracy; they are fine for visualization, not as a basis for manufacturing.

Image-based photogrammetry fails on plain single-color, shiny, transparent and very dark surfaces and on thin features. Such surfaces have to be sprayed or given a pattern. Blurred images, insufficient overlap between shots, changing light or a moving object show up as noise and holes in the data.

In your request, state whether you need point coordinates with a declared accuracy or a textured model for presentation. Ask what accuracy the provider guarantees for your object size and how they verify it. Also check whether targets may be stuck onto the surface and whether marks left after removal are acceptable.