Is a 3D scan the same as a CAD file?

No. A scan typically captures geometry; a CAD model often encodes designed features and constraints.

Does every scanner use a laser?

No. Structured light and other sensing approaches are also used.

Can scanning capture hidden internal details?

Ordinary external surface scanning cannot directly observe inaccessible interiors; that requires an appropriate different measurement method or additional access.

AI Glossary / Computer vision and multimodal AI / Advanced

3D scanning

Definition

3D scanning is the process of capturing an object's or environment's geometry as three-dimensional measurements. Many systems produce point clouds that can be processed into meshes or other models. Scanners use different sensing principles, including projected light, laser triangulation and distance measurement, with performance depending on the method and conditions.

What it means

A scan provides spatial observations of surfaces that the sensing system can measure. It is not automatically a complete digital replica. Some regions may be hidden, poorly measured or reconstructed later by software. A scan also differs from a finished computer-aided design model: measured points do not inherently describe design intent, material properties or how a part should be manufactured. The right interpretation depends on whether the task is visualization, inspection, documentation or another use.

How it works

Set up and calibrate the system, capture geometry from useful viewpoints and register multiple captures into a shared coordinate frame. Processing may filter noise, fuse observations and reconstruct surfaces. Structured-light systems project patterns and infer geometry from their observed deformation using calibrated geometry. Other scanners measure distance using different optical principles. Surface reflectance, angle, range, occlusion and calibration can affect results. Validation should use suitable references and conditions relevant to the application, rather than relying only on nominal point count or a smooth rendered model.

Three practical examples

A manufacturer scans a component to compare its measured surface with a reference design, using an inspection method appropriate to the required tolerances.
A facilities team captures a room or plant area to document visible geometry before planning changes, while checking critical dimensions independently where needed.
A designer scans a physical prototype as a starting point for a digital model, then rebuilds features that require explicit dimensions and design constraints.

Think of it this way

Think of collecting a dense set of spatial survey notes. The notes can be assembled into a useful representation, but density does not guarantee correctness. Repeatedly recording a point in the wrong place does not improve its accuracy. Likewise, a scanner's millions of points can still contain systematic error or leave a small but important feature unobserved.

Worked example

In an original inspection illustration, a reference distance is 100.00 millimeters. Three scan-based measurements are 100.20, 100.22 and 100.18 millimeters. Their mean is 100.20 millimeters and their range is 0.04 millimeters. The measurements are closely grouped, but their mean differs from the stated reference by 0.20 millimeters. This demonstrates why repeatability and agreement with a reference are different questions. A real assessment would also consider reference uncertainty, alignment, environmental conditions and the measurement procedure. The calculation does not establish an acceptance decision for any particular part or scanner.

Illustrative case study

A fictional fabrication workshop scans a replacement housing whose original drawings are incomplete. The scan captures broad surfaces well but leaves a deep recess and reflective edge poorly represented. The team marks those regions as uncertain, obtains additional measurements and constructs a design model with explicit dimensions. It checks the final design against the physical interfaces rather than treating the first mesh as production-ready. The workflow distinguishes captured geometry from engineering decisions made during reconstruction. This case illustrates a possible use of scanning, not evidence that an arbitrary scan is accurate enough for manufacturing.

Common misconceptions and mistakes

Resolution, point density, precision and accuracy are related but different properties. A high-density point cloud can still be biased or noisy. Registration error can also distort a model assembled from individually good scans. Another mistake is interpreting filled holes as measured surfaces. Reflective, transparent or very dark regions can challenge optical methods, and the specific difficulty depends on the sensor. Finally, a device's specification under one test condition should not be assumed to describe every material, distance and viewing angle in a real task.

Business use

3D scanning can shorten some documentation and modeling tasks and support inspection where its performance is suitable. A useful business evaluation includes capture, setup, processing, cleanup and verification time, not just acquisition speed. Define the smallest relevant feature and the accuracy needed for the decision before choosing a workflow. Data formats, coordinate systems and traceability also matter when scans move between teams. The deliverable should distinguish raw observations, processed surfaces and any manually reconstructed geometry so downstream users understand what they are relying on.

Check your understanding

What did the small measurement range show?

Answer: Good agreement among those repeated measurements.

What did the 0.20-millimeter mean difference show?

Answer: A discrepancy from the stated reference in the example.

Why should a filled recess be labeled carefully?

Answer: Software completion is not the same as direct measurement.

Related terms

Point cloud
structured light
laser triangulation
registration
surface reconstruction
dimensional metrology

Learn next

Compare raw and processed scan data for a simple object. Identify occluded regions and alignment choices. Then measure an independent reference and separate repeatability from accuracy before judging whether the workflow meets its intended purpose.

Sources

Rachakonda, Muralikrishnan and Sawyer, Sources of Errors in Structured Light 3D Scanners: https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=927473 NIST, Performance Evaluation of Terrestrial Laser Scanners—A Review: https://www.nist.gov/publications/performance-evaluation-terrestrial-laser-scanners-review The measurement arithmetic and fabrication scenario are original educational illustrations. They do not specify universal scanner tolerances or acceptance criteria.

Previous: 3D reconstruction from multiple images Next: 8-point algorithm