Learn object scanning

How Object 3D Scanning Works: Accuracy, Standards, and the Full Pipeline

By the ScanningAndModeling team · Last updated July 16, 2026

Object 3D scanning measures the surface of a physical thing, from a fishing lure to a hundred-foot vessel, and turns it into digital geometry you can inspect, reproduce, or reverse engineer. Three technology families do this: photogrammetry, structured light, and metrology-grade laser. They are not interchangeable, and the honest way to choose between them is not by marketing adjectives but by how each one gets its accuracy. This guide explains the technologies, what accuracy really means in microns and thousandths, the actual standards that govern it, the surfaces that fight back, and the full pipeline from first point to finished deliverable.

Our aim here is to give you the real thing, not a brochure. By the end you should be able to tell a metrology-grade claim from a hopeful one, know which capture method fits your part, and understand exactly what happens between the scan and the file you receive.

What are the three ways to 3D scan an object?

There are three real capture families at object scale. The most important distinction between them is whether their accuracy is measured directly or derived by computation, because that single difference decides whether a scan is trustworthy for a manufactured part.

Photogrammetry

Photogrammetry reconstructs 3D geometry from many overlapping photographs. There is no laser and no projected pattern, just cameras and software that triangulate common points across images. Its accuracy is derived: computed from image analysis, so it depends on lighting, lens, texture, and how well the software solves the scene. Typical accuracy runs roughly 0.5 to 2 mm, and because the result is calculated rather than measured, scale can drift and relative error can exceed ten percent in poor conditions. Photogrammetry shines where color and texture matter and absolute precision does not: archaeology, film and game assets, terrain, and large scenes. It is the honest choice when you want a faithful-looking model at low cost, not a metrology answer. See the photogrammetry versus 3D scanning comparison for the tradeoffs in detail.

Structured-light scanning

A structured-light scanner projects a known pattern of stripes or fringes onto the object and reads how that pattern deforms across the surface with one or more cameras. Because the pattern is known, the geometry is recovered with far more metric stability than photogrammetry, roughly an order of magnitude better or more, and it works on plain, featureless surfaces that defeat photogrammetry. This is the workhorse for reverse engineering mechanical parts, dental and medical work, jewelry and small parts, and production quality control. It captures a whole area at once, which makes it fast over complex shapes. A side-by-side of the three technologies lays out where each is appropriate.

Metrology laser scanning

A laser scanner measures distance directly. A laser range measurement is a physical distance, not a computed guess, which is why laser wins on raw depth accuracy. Professional line and spot systems reach around 0.05 mm or better, and for verifying a part against tight GD&T tolerances, a laser scanner or a coordinate measuring machine is the appropriate tool. The structured-light versus laser breakdown covers where each earns its place.

How accurate is 3D scanning, really?

Accuracy is where honest vendors separate from optimistic ones, so here it is in plain numbers. For inspection-grade work, structured-light and laser scanners typically resolve point accuracy in the range of 0.02 mm to 0.1 mm, which is 20 to 100 microns. To translate for a shop floor: 0.025 mm is about 0.001 inch, one thousandth. Machinists think in thousandths, metrologists think in microns, and a real scanning service speaks both. When someone quotes you accuracy, ask which literacy they are using and under what conditions.

The deeper honesty point is the difference between stated accuracy and accredited accuracy. Stated accuracy is the spec-sheet number a manufacturer prints under ideal lab conditions. Accredited accuracy is a number verified by an ISO/IEC 17025 accredited lab under a defined test protocol. They are not the same, and a page that respects the distinction is telling you the truth. Creaform, a metrology maker, writes clearly on the difference between stated and accredited accuracy. On top of that, your real-world result depends on part size, surface finish, temperature, and the capture setup itself, so a single-number promise divorced from those conditions should always be read skeptically.

Which standards govern scanner accuracy?

Two families of standards actually exist for optical 3D measuring systems, and naming them correctly is how you tell an expert from a marketer.

• VDI/VDE 2634. A German guideline for optical 3D measuring systems. Part 2 covers area-scanning (structured-light) systems, and Part 3 covers multiple-view area-scanning systems. A scanner certified to VDI/VDE 2634 Part 3 has been tested for accuracy, repeatability, and reliability against a defined protocol.
• ISO 10360-13. Part of the ISO 10360 series that specifies performance requirements and acceptance and re-verification procedures for coordinate measuring systems. ISO 10360-13 specifically addresses optical 3D (structured-light) systems. NIST research found that ISO 10360-13 length tests give better sensitivity to model parameters than VDI/VDE 2634-2, documented in the NIST study on VDI/VDE 2634-2 and ISO 10360-13 performance evaluation.

Credible scanners are verified against these protocols by an ISO/IEC 17025:2017 accredited lab. If a scanning claim cannot point to VDI/VDE 2634 or ISO 10360-13, treat the accuracy number as marketing until proven otherwise.

What size objects can be scanned?

The range is enormous, and the method changes with the size. Small parts, down to roughly a millimeter of feature, a fishing lure, a bracket, a medical stent, are best captured on a bench with structured light or laser, and are usually mailed in and scanned in a controlled setup. Large objects, a classic-car body, a boat hull, a piece of production tooling, or a full vessel up to a hundred feet, are captured on site with portable equipment, often blending laser scanning with photogrammetry to cover form and detail. The principle holds at every scale: what the scanner never sees cannot appear in the result, so coverage is planned backward from the deliverable. For the largest objects, work that would take weeks of manual measurement is done in days at a fraction of the cost.

Which surfaces are hard to scan, and what do we do about it?

Optical scanning reads light bouncing off a surface, so surfaces that misbehave with light are the real-world challenge. Being honest about them up front is part of scoping a job correctly.

• Shiny and polished surfaces reflect the laser or projected pattern away from the cameras, leaving gaps or noise.
• Dark and matte-black surfaces absorb light, so too little returns to measure cleanly.
• Transparent and translucent surfaces let light pass through instead of reflecting it, so the scanner cannot find the true surface.

The standard, well-understood fix is a thin, temporary matte coating, a fine scanning spray that dulls the surface just long enough to capture it and then wipes or sublimates away. On critical parts we account for the coating thickness in the accuracy budget. The point worth knowing as a buyer: these surfaces are scannable, they just require the right preparation, and a vendor who pretends a mirror-polished part scans perfectly untreated is not being straight with you.

What is the full pipeline from scan to deliverable?

Here is the whole arc, root to finished product. Each step adds value, and the expensive value is at the far end, not the near end.

The point cloud is the dense set of measured points. It is converted into a watertight mesh, a closed shell of triangles that represents the part, as described by reverse-engineering bureaus like Exact Metrology. In reverse engineering the mesh becomes a true CAD model with editable features, planes, cylinders, NURBS surfaces, and boolean operations, as Verisurf documents. That reconstruction is hybrid: prismatic shapes are extracted as parametric features while organic geometry is fitted with freeform NURBS surfaces, a split explained well by this point-cloud-to-CAD overview. The output is STEP, IGES, or Parasolid for production.

The reason this pipeline matters to a buyer is that the effort is heavily loaded at step three. Capturing points takes minutes to hours. Turning a noisy mesh into a clean, tolerance-correct parametric solid is skilled human labor, commonly several hours to well over ten depending on complexity. That is why "the scan is the easy part" is not a slogan but a description of where the cost lives. We treat that distinction in full in our companion guide, mesh vs CAD, what a scan actually gives you.

What is as-built versus design-intent reverse engineering?

When the mesh becomes CAD, there are two honest philosophies for the rebuild. As-built reproduces the part exactly as it physically is, warts and wear included. Design-intent reconstructs the part as it was meant to be: clean symmetry, round-number dimensions, true right angles. Laser Design, a bureau with decades of reverse-engineering work, frames this well in its reverse-engineering overview. Which one you want depends on the goal: replace a worn part and you usually want design-intent so the replacement is correct; document an existing part exactly and you want as-built. We ask at scoping, because it changes how the model is built.

How does 3D scanning inspect a part against tolerances?

Beyond reproducing a shape, scanning is a powerful inspection tool. Full-field inspection overlays the scan against the nominal CAD model and renders a color deviation map, a heat chart where green means in tolerance, red means high, and blue means low. Reading the map is itself a skill: a uniform offset across the whole part points to global shrinkage, a smooth gradient suggests a thermal or cooling imbalance, and localized peaks flag mold damage or ejector interference, as detailed in this explanation of full-field deviation analysis. Alignment comes first, usually a best-fit or N-point alignment, then a datum alignment for formal GD&T, as covered in this piece on first-article inspection with 3D scanning. The software can even act as a virtual coordinate measuring machine, measuring the digital part and generating a report as if it had been probed on a real CMM.

The honest limit belongs here, not buried. Scanning captures a full surface fast at 0.02 to 0.1 mm, which is superb for whole-part deviation and for most features. But for the tightest prismatic tolerances and pass-fail GD&T on a critical bore, a touch-probe CMM remains the safer method, a point made plainly in this CMM versus 3D scanner comparison. We will tell you when your part is a scanning job and when it genuinely needs a CMM, rather than overselling one tool for every job.

How we run an object scan, walked through

You do not need any of the vocabulary above to get the right result. Translating your goal into the correct method is our job. Here is how a project actually runs.

Step 1: we scope from the goal, not the jargon

You tell us the object and what you intend to do with it, machine it, mold it, inspect it, reproduce a discontinued part, print a copy. From that we choose the capture method, the accuracy target, and the deliverable. You describe the outcome. We specify the scan.

Step 2: capture, on the bench or on site

Small and medium parts are typically mailed in and captured on a bench with structured light or laser under controlled conditions. Large or immovable objects are captured on site with portable equipment. Difficult surfaces are prepared with temporary matte spray. Coverage is planned backward from the deliverable so nothing the model needs is missed.

Step 3: mesh, reverse engineer, and check

The point cloud is cleaned and meshed, then, if you need CAD, rebuilt into a parametric solid feature by feature, as-built or to design-intent as you specified. The result is checked back against the scan with a deviation map so the model is verified, not assumed.

Step 4: deliver in the format your workflow opens

You receive the deliverable your job requires: a watertight STL or OBJ mesh for printing and reference, or a parametric STEP file, with IGES, Parasolid, a native format, or a dimensioned 2D drawing as needed, plus an inspection report when the job is a comparison. We confirm the target format up front so files open cleanly on your side.

Questions buyers actually ask

Which scanning technology is most accurate?

Metrology laser and structured light both measure accuracy directly and reach the tens of microns, while photogrammetry derives accuracy by computation and typically lands in the half-millimeter to two-millimeter range. For manufactured parts you want a measured method. Photogrammetry is best where color and cost matter more than precision.

What does 3D scanning accuracy in microns or thousandths mean?

A micron is a thousandth of a millimeter. Inspection-grade scanning typically resolves 20 to 100 microns, and 0.025 mm is about one thousandth of an inch. Machinists speak in thousandths and metrologists in microns, and both describe the same physical tolerance.

Can you 3D scan shiny, black, or transparent objects?

Yes, with preparation. Reflective, dark, and transparent surfaces confuse optical scanners, so we apply a thin temporary matte spray that lets the scanner read the surface and then wipes or sublimates away. On critical parts we account for the coating in the accuracy budget.

Is 3D scanning as accurate as a CMM?

For full-surface deviation and most features, scanning is excellent and much faster. For the tightest prismatic tolerances and pass-fail GD&T on critical features, a touch-probe CMM is still the safer method. We tell you which one your part actually needs.

How big an object can you scan?

From a fishing lure to a hundred-foot vessel. Small parts are captured on a bench, usually mailed in. Large objects are captured on site with portable equipment, often blending laser and photogrammetry. The method scales with the object.