Structured-light scanners
Optical triangulation, acquisition and registration controls, performance specifications, and limits of inspection meshes.
Tags: inspection, dimensional-metrology, optical-measurement, scanning
On this page
Measurement principle
A structured-light scanner triangulates surfaces from projected patterns and calibrated camera observations. NIST: paper and publication record, full paper
Illustrative sampling example: assume a scan samples a region at approximately 0.10 mm spacing. Exporting a mesh with nominal 0.02 mm edges does not add new optical observations between those samples. Interpolation may improve representation for a downstream tool, but the smaller mesh spacing does not demonstrate 0.02 mm measurement accuracy. These assumed values illustrate the distinction, not a specification for a named scanner.
The useful output depends on the task. A reverse-engineering mesh, a nominal-to-actual surface comparison, and a dimensional acceptance report need different processing and validation. Start with the feature, tolerance, surface condition, and required coordinate system rather than a desired polygon count.
Follow the data through the process
| Stage | Output | Question to resolve |
|---|---|---|
| Acquisition | Surface samples from individual views | Was the required surface actually observed with adequate signal and coverage? |
| Registration | Views transformed into a common frame | Is their relative position sufficiently constrained and independently checked? |
| Fusion and meshing | A combined surface representation | Which regions were measured, filtered, smoothed, or filled? |
| Inspection alignment | Measured geometry located relative to nominal geometry | Does the alignment implement the required datums or another explicitly chosen comparison? |
| Evaluation | Fitted features, deviations, and decisions | Are sampling, fitting, uncertainty, and acceptance rules appropriate? |
Treat a visually complete model as a processed representation. It is not proof that every triangle came from a directly observed surface.
Measuring volume and a bounded model example
ZEISS's undated ATOS Q product page lists 8M and 12M versions, measuring areas from 100 × 70 to 500 × 370 mm², and 490 mm working distance. Its table lists point-distance ranges of 0.04–0.15 mm for 8M and 0.03–0.12 mm for 12M. ZEISS describes interchangeable lenses for different measuring volumes. Those figures apply to the documented configurations; do not combine the finest spacing with the largest measuring area unless the applicable configuration explicitly supports it. ZEISS: ATOS Q specifications
Point distance describes sampling. It does not establish an equal dimensional error limit or guarantee reliable measurement of a feature of that width. Obtain the exact performance specification, test method, measuring volume, and applicable environmental conditions for the supplied system. A small field may provide denser sampling but require additional views to cover the part, making registration performance relevant to the final result.
Surface visibility and preparation
Optical access is an acquisition requirement. Plan views that expose the required surfaces and check coverage before disassembling the setup. A deep recess, narrow gap, or hidden face may remain unavailable despite a dense exterior scan. A missing region should trigger another acquisition or a separately identified modeling operation.
Artec Studio 19 guidance recommends powder or anti-glare coating for transparent, reflective, or black objects when needed. That recommendation is application-dependent: coating must be permitted on the part, and its dimensional effect must be assessed when relevant to the tolerance. Record the surface treatment and retain evidence of the condition being measured. Artec: Studio 19 scanning guide
Qualify exposure and surface preparation on representative material and finish, not only on a convenient matte artifact. Stabilize the component and fixture during the acquisition sequence. Check whether motion, changing illumination, or surface response produces missing or inconsistent observations. Repeated passes that share the same systematic limitation do not establish correctness by agreeing with one another.
Registration is different from inspection alignment
Registration joins individual scans. Its constraints may come from overlapping geometry, texture, reference targets, or another system-specific arrangement. Establish which method is active, where overlap is needed, and whether the geometry supplies enough information to constrain motion. Repetitive or nearly featureless shapes merit particular scrutiny. Check joined regions for inconsistent surfaces and assess independent reference dimensions across the assembled scan.
Inspection alignment positions the measured object relative to its reference model. SHINING3D Inspect 2026, documentation version 1.5.0.9, distinguishes best-fit, datum, 3-2-1, RPS, and other alignments. Its best-fit method minimizes overall mean distance deviation; its datum alignment instead uses corresponding datums to align measured and reference geometry. These solve different problems. SHINING3D: alignment documentation
For drawing-based inspection, apply the required datum definition and constraints. A freely optimized best fit can redistribute deviations and change a color map without changing the underlying samples. State the alignment and selected regions so the reader knows what the displayed deviations mean. A green map is meaningful only with its scale, filtering, comparison method, and acceptance requirements.
Meshing and derived geometry
Artec Studio 19 separates registration from fusion and offers hole-filling choices including no filling, watertight, and radius-based filling. It also provides mesh simplification and editing tools. For dimensional work, preserve the original observations and distinguish synthesized regions from acquired ones. Smoothing, simplification, and filling need task-specific evaluation before their outputs are used as measurement evidence. Artec: processing workflow
Calibration, verification, and uncertainty
ISO 10360-13:2021, published September 2021, addresses acceptance and reverification of optical 3D coordinate measuring systems for manufacturer-stated length performance. Its public scope restricts the scanned object's surface characteristics to a cooperative range. Passing such tests therefore needs to be distinguished from validating an application involving different gloss, color, or geometry. ISO: ISO 10360-13 scope
NIST's 2019 work similarly cautions that artifact-based specifications may not reveal all errors encountered with different real-world objects. Use the applicable test specification and calibrated artifacts, then assess representative application conditions separately. Do not carry forward the paper's historical statement about standards under development as the current standards situation. NIST: sources of scanner errors
An August 2023 ORNL study explicitly investigated repeatability and reproducibility while including part position and orientation relative to the scanner. That is a useful design for local studies: include reorientation and repeated acquisition, rather than merely recomputing a dimension from the same mesh. Its results describe the studied system and artifact, not a universal scanner capability. ORNL: artifact-based performance study
Evaluate calibration, surface effects, registration, temperature, fixture deformation, sampling, and processing in the uncertainty assessment appropriate to the measurand. Keep repeatability separate from systematic error. Report scanner/lens configuration, software version, surface preparation, coverage limits, registration and datum alignment, processing settings, calibration status, results, and required uncertainty and decision rule.