Structured Light Scanner

What is a Structured Light Scanner

A structured light scanner is an optical, non-contact 3D coordinate measuring system. Its core idea is triangulation: a projector casts a controlled pattern of light onto the part, one or two cameras observe that pattern from a known angle, and software computes a 3D coordinate for every pattern point by intersecting the projector ray with the camera line of sight. The fixed geometric baseline between projector and camera, established during calibration, converts an apparent shift of the pattern in the camera image into a depth value. Unlike a touch-probe coordinate measuring machine that records one point per contact, a structured light scanner records a full field of points in a single capture sequence.

Physically, a typical system has three functional blocks: (1) the projection unit, today usually a digital light projector based on a DLP micromirror chip illuminated by a blue or white LED, which can display any computed pattern; (2) the imaging unit, one or two industrial cameras (commonly 5 to 12 megapixels) with a narrow band-pass filter matched to the projector wavelength; and (3) the processing and software stack that decodes the pattern, computes the point cloud, registers multiple scans, and exports a polygon mesh. When two cameras straddle the projector in a stereo arrangement, the system gains a second independent measurement of every point, which improves robustness and lets the scanner self-check calibration.

The technique grew out of moiré and fringe projection profilometry developed through the 1970s and 1980s. The arrival of programmable DLP projectors in the late 1990s freed the field from fixed gratings, allowing software to project Gray-code sequences and phase-shifted sinusoidal fringes on demand. The commercial breakthrough for industrial metrology came from German firm GOM, whose ATOS area-scanning systems made VDI/VDE 2634-certified optical inspection routine in automotive body shops. The later move from white to blue LED projection, and the addition of photogrammetry-referenced multi-scan workflows, pushed accuracy and shop-floor usability to where structured light now competes directly with tactile CMMs for many freeform and sheet-metal parts.

It helps to place the structured light scanner among neighboring instruments. A laser line scanner sweeps one or several laser stripes and is more tolerant of glossy and dark surfaces; a laser tracker measures large volumes at long range with a retroreflector; a confocal or chromatic displacement sensor measures a single high-precision point or profile at micron scale; and a tactile CMM remains the reference for the tightest single-feature tolerances. Structured light sits in the middle: very high point density over a bounded volume, excellent for whole-surface deviation maps, sensitive to surface optics, and dependent on careful calibration and registration for global accuracy.

Four engineering metrics decide whether a structured light scanner suits a job: accuracy (stated against a named standard and measuring volume), resolution and point spacing, measuring volume and standoff, and surface cooperativity (how well it handles real material finishes without spray). These four, combined with certification and service availability, drive both capability and total cost of ownership over the instrument lifecycle.

Projection Types and Configurations

Structured light scanners differ first by light source and second by mechanical configuration. The light source sets ambient-light tolerance and fine-detail contrast; the configuration (fixed versus handheld, single versus dual camera) sets accuracy class, portability, and how the scanner handles large or hard-to-reach parts. The table below compares the principal options on the parameters that matter at purchase.

Blue LED projection has become the default for industrial metrology. The narrow emission band near 450 to 460 nm pairs with a matched band-pass filter on the cameras so that almost all ambient light is rejected, letting the scanner work in a lit factory or under daylight without the pattern washing out. The short wavelength also forms sharper fringe edges, which raises contrast and lowers measurement noise on fine features. ZEISS reports that the shift from white to blue light is the main reason most high-end units now run blue.

White LED projection covers the full visible spectrum and was the original industrial approach. It captures true surface color well, which suits some heritage and consumer scanning, but it cannot be spectrally filtered as cleanly, so it is markedly more sensitive to room lighting and generally restricted to controlled environments. Many vendors have retired white-light metrology lines in favor of blue.

Fixed dual-camera systems mount projector and two cameras on a rigid bar, calibrated to a fixed geometry, usually on a tripod facing the part or a rotary table. The stereo pair gives two independent depth estimates per point, which raises accuracy into the single-digit-micron range on small volumes and enables continuous self-monitoring of calibration. These are the reference tools for first-article inspection and tool validation.

Handheld portable scanners trade some accuracy for the freedom to walk around a car body or an aircraft panel. They keep alignment by tracking reference targets stuck to the part or surrounding frame, or by registering on geometry and texture in real time. Many recent handheld units are hybrids that can switch between structured-light area capture and laser-line capture so the operator can use whichever mode the surface allows.

Pattern Coding Methods

The pattern a scanner projects is not arbitrary: it must let software assign every observed image point to a unique projector coordinate, a problem called the correspondence problem. The coding scheme chosen sets the trade-off between speed, robustness to noise, and final resolution. Three families dominate, and most metrology scanners blend two of them. The table compares the main coding methods on the properties an engineer should weigh.

Binary coding projects a sequence of black-and-white masks, each splitting the field into finer regions, so that the on/off sequence at any pixel encodes a unique column number. It is conceptually simple and only ever stores two intensity values per pixel, but the finest stripes are sensitive to defocus and surface reflectivity, and the method needs the part to hold still through the whole sequence.

Gray code is a refinement of binary coding in which successive patterns differ by only one bit, so a misread at a stripe boundary causes at most a single-step error rather than a large jump. This makes Gray code notably more robust to decoding errors than plain binary, which is why it is the standard coarse-correspondence layer. Its resolution is still limited by stripe width, so on its own it cannot reach sub-pixel precision.

Phase-shifting projects continuous sinusoidal fringes and shifts them by a known fraction of a period (commonly three or four steps). Because intensity varies smoothly, software solves for a sub-pixel phase value at every pixel, which gives very high spatial resolution and data density. The catch is periodicity: the phase repeats every fringe, so the raw result is ambiguous (wrapped) and must be unwrapped to an absolute coordinate.

Combined Gray-code plus phase-shift resolves that ambiguity by using the Gray-code or binary layer to assign a unique period number to each fringe and the phase-shift layer to interpolate finely within each period. This pairing delivers both high accuracy and strong robustness against ambient light, at the cost of projecting ten to twenty frames per scan, so it suits static metrology rather than fast moving capture. Single-shot methods project one coded dot or speckle pattern and decode depth from a single frame, trading resolution for the ability to capture moving scenes; they appear in real-time and consumer depth sensors more than in high-accuracy industrial inspection.

Surfaces, Materials, and Limits

The single biggest limitation of structured light scanning is not the electronics but the optics of the target surface. The method assumes a cooperative surface that scatters projected light diffusely back to the cameras. Real industrial materials often violate that assumption, and understanding when they do is the difference between a clean scan and hours of rework. Both VDI/VDE 2634 and ISO 10360-13 explicitly restrict their stated accuracy to surfaces whose gloss and color stay within a cooperative range, which is the standards bodies acknowledging this physical reality.

Shiny and specular surfaces (polished metal, chrome, glossy paint, mirrors) reflect the projected pattern in a focused, mirror-like way rather than scattering it. The reflected light either misses the cameras entirely, leaving gaps, or arrives as intense glare that overexposes pixels and produces noise and floating artifacts in the point cloud. Dark surfaces absorb most of the incident light, so little returns to the cameras and the scanner records weak signal, holes, or noisy data. Transparent and translucent surfaces (glass, clear plastic, some ceramics) let the pattern pass through or scatter below the surface, so the cameras never see a sharp pattern on the true surface.

The common engineering remedy is a temporary matting coat of 3D scanning spray that forms a thin, uniform, diffuse white film. Conventional titanium-dioxide sprays must be cleaned off afterward, while vanishing or sublimating sprays evaporate within minutes to hours and leave no residue. The caveat is metrological: the film has thickness, typically a few microns, which adds a small systematic offset to the measured surface. For loose tolerances this is negligible; for tight inspection it must be accounted for or avoided. Other mitigations include polarizing filters to cut specular glare, HDR multi-exposure scanning that combines short and long exposures to capture both bright and dark regions, and lowering projector intensity for very reflective parts.

The table below is a quick-reference lookup for common surface types and the recommended approach. It is a starting point only; on critical parts, validate the approach against a calibrated artifact of similar finish before trusting production data.

Geometry imposes a second class of limits. Deep narrow cavities, sharp internal corners, and undercuts can shadow either the projector ray or the camera line of sight, leaving data holes that only re-orientation or a different instrument can fill. Standoff distance also matters: as the part moves away from the nominal standoff, the projected pattern spreads and defocuses, degrading resolution, which is why each lens or measuring volume has a defined working distance and depth of field.

Key Specification Parameters

Reading an optical scanner datasheet is a skill, because vendors quote the same capability in different ways. Eight parameters truly drive a metrology selection: accuracy (against a named standard), resolution and point spacing, measuring volume and standoff, points per scan and scan speed, camera resolution, calibration and stability, certification, and reference-target or photogrammetry support. The table below compares the headline metrology specifications a buyer should normalize across quotes.

Accuracy is meaningful only with its context. Reputable vendors state it as a probing or length-measurement error measured per VDI/VDE 2634 Part 2 (single view), Part 3 (multiple view), or ISO 10360-13:2021, on a defined measuring volume, using calibrated spheres or sphere bars. Some figures include a length-dependent term, for example a fixed value plus a per-metre coefficient that grows with part size. A bare accuracy number with no artifact, volume, or standard named cannot be compared between brands, so always request the calibration certificate.

Point spacing (also called point distance) is the average gap between adjacent measured points and sets the smallest feature you can resolve. As a working rule, point spacing should be roughly one third to one fifth of the smallest detail you must capture. Published examples illustrate the lens trade-off: the ZEISS ATOS Q lists point distance around 0.04 to 0.15 mm on its 8-megapixel sensor and 0.03 to 0.12 mm at 12 megapixels, while the ZEISS GOM Scan 1 lists point distances of about 0.037, 0.060, and 0.129 mm across its three measuring-area lenses.

Measuring volume and standoff define the part size captured per scan and the working distance. They trade against point spacing: a smaller volume concentrates the same camera pixels onto less area for finer points, and a larger volume covers a big part faster but coarser. Metrology systems often ship interchangeable lenses so one scanner can serve small and large volumes. As a published handheld example, the Creaform Go!SCAN 20 has a scanning area near 143 by 108 mm at a standoff around 380 mm, while the Go!SCAN 50 covers about 380 by 380 mm at roughly 400 mm standoff.

Points per scan and scan speed describe data density and throughput. Industrial area scanners commonly capture up to several million points per scan; the ZEISS ATOS family is documented capturing up to 12 million points per scan, and the GOM Scan 1 up to about 6 million. Handheld units quote a measurement rate instead, for example the Creaform Go!SCAN line capturing on the order of 1.5 million measurements per second. Camera resolution (commonly 5 to 12 megapixels per camera) ultimately bounds both point spacing and accuracy for a given measuring volume.

Calibration, stability, and certification determine whether quoted accuracy survives shop-floor use. Look for self-monitoring calibration, a documented temperature operating range, and a calibration certificate naming VDI/VDE 2634 or ISO 10360-13, ideally backed by an accredited laboratory such as a CNAS- or DAkkS-accredited facility. Reference-target and photogrammetry support decides how well the scanner holds global accuracy on large parts: coded targets and an external photogrammetry camera let many local scans lock into one reference frame instead of accumulating registration drift.

Selection Decision Factors

To turn the preceding five chapters into a specific model choice, follow the decision sequence below. Most selection mistakes come not from one wrong answer but from deciding accuracy class before the part, surface, and volume are pinned down. These eight steps double as a fixed RFQ template.

1. Part size and measuring volume: Define the largest and smallest parts in scope. This sets the measuring volume and whether you need a single fixed volume or interchangeable lenses. Oversizing the volume wastes resolution; undersizing forces excessive stitching on large parts.
2. Required accuracy and the standard behind it: Separate reverse engineering and styling (where 0.05 to 0.1 mm is fine) from dimensional inspection against tolerance (where 0.01 to 0.03 mm and a VDI/VDE 2634 or ISO 10360-13 certificate are needed). Demand the artifact, volume, and standard with every accuracy figure.
3. Smallest feature and point spacing: Identify the smallest hole, edge, or fillet you must resolve, then require point spacing roughly one third to one fifth of it. This often decides camera resolution and lens choice.
4. Surface and material: Audit real surfaces against Chapter 4. If parts are shiny, dark, or transparent and spray is not acceptable, weigh blue light with polarizing and HDR features, or consider a laser-line or tactile alternative.
5. Fixed versus handheld: Lab inspection of repeated parts favors a fixed dual-camera system on a tripod or rotary table. Scanning large assemblies, vehicles, or fixed installations favors a handheld or hybrid unit with target or photogrammetry referencing.
6. Global accuracy strategy for large parts: If parts exceed one measuring volume, plan the reference scheme now: coded targets, an external photogrammetry camera, or both, to stop registration error from accumulating across many scans.
7. Software and workflow fit: Confirm export formats (STL, mesh, point cloud), inspection and GD&T reporting, CAD-compare and color-deviation maps, and integration with your existing metrology or CAD chain. Software often determines daily productivity more than raw hardware specs.
8. Total cost of ownership (TCO): Purchase price plus annual calibration, consumables (scanning spray and targets), software maintenance, operator training, and downtime. A low entry price with weak service or no accredited calibration path can cost more across a multi-year inspection program.

One frequently overlooked dimension is serviceability and calibration access: local calibration laboratories, recertification turnaround, spare-projector and camera availability, and firmware and software update cadence. ZEISS GOM, Creaform (Hexagon), Artec 3D, KEYENCE, and Carl Zeiss maintain global service and certified calibration; China-based suppliers such as SHINING 3D, Revopoint, and Scantech increasingly offer CNAS-accredited calibration and VDI/VDE 2634 certificates at lower price points. For a long-running inspection cell, confirm the recertification path before purchase, not after.