3D documentation of an examination with an iPhone: scale, uncertainty and integrity

Photogrammetry and LiDAR

Photogrammetry builds a 3D model from a series of overlapping images. On its own, however, it does not know how large the object is: the model is to scale only when something supplies the scale. On iPhones with LiDAR every frame has depth, the direction of gravity and a metric camera trajectory from which the scale is derived. Without LiDAR, the ARKit trajectory and a calibration target in the frame supply it. For forensic use it matters that the model is built only from captured data, with no AI-completed geometry.

The scale must be verified

A scale from a single source can be wrong. Independent routes are therefore compared: the metric camera trajectory, the calibration target found in the model and a reference length measured in the model. If the scale sources disagree, the target decides and the uncertainty grows. A reference shorter than 290 mm cannot resolve a 1 % error, so body length calls for a 500 to 1000 mm reference, such as a measuring tape.

Uncertainty ± U (k = 2)

Every measurement in the model carries an expanded uncertainty U with a coverage factor k = 2, roughly a 95 % level of confidence under the GUM metrology standard. It combines the uncertainty of the scale, the precision with which points are placed on the surface and the resolution of the model. A measurement without a stated uncertainty is an estimate, not a measurement.

Documentation classes A to D

The 3D documentation certificate states how many independent routes support the scale and with what result. Class A is metrologically verified (LiDAR depth, scale corrected by a trajectory fit and an independent check able to resolve 1 %). Class B is metric but lacks a sufficient independent check. Class C has only an approximate scale and class D no known scale. The worst of the available checks decides.

Integrity: SHA-256 and signing

Documentation is useful only if it can be shown not to have changed. The .f3d package carries a SHA-256 of every file, so the recipient sees on import whether anything is missing or was altered. Processing on a Mac adds a reconstruction record (tool, version, stages, duration) and can sign the output package with the computer's key.

Limits

• Not a certified or metrologically traceable measuring instrument; supplementary, illustrative documentation.
• Accuracy depends on light, movement, shiny and dark surfaces and on how completely the object is covered by images.
• After recomputing a model, a new reference measurement and a new certificate are needed.

How aPrehliadky and Fotograme3ka 3D handle it

aPrehliadky for iPhone and iPad captures with LiDAR, computes the model on the device through RealityKit Object Capture, measures distance, surface path, area and volume with an uncertainty ± U (k = 2) and issues a PDF report, a class A to D certificate and an .f3d package with SHA-256. Fotograme3ka 3D on an Apple Silicon Mac recomputes the package at higher detail and issues a PDF/A-2b report with a signed package.