Digitizing Cultural Heritage: From 3D Model to Useful Visitor Service

Decide the end use before you capture

The most common mistake is to buy a scanner or run a mass photo campaign before deciding what the result is for. The model then sits on a server and serves neither conservators nor visitors. Split the two goals. Documentation and preservation demand accurate geometry, faithful colour and structured metadata. A visitor service demands a fast-loading copy, storytelling and augmented reality.

Research on quality in 3D digitisation of tangible cultural heritage, which ICOMOS and its documentation committee contributed to, makes the same point from the evidence side: the complexity of a digitisation project depends on the equipment, the placement and lighting of the object and the texture of its material — and the right approach changes with the intended use. Turn that into a working habit: write down who will view the model and where, which details are critical, and what weight and format your publishing channel allows, before you commit budget and a contractor.

A single capture session can feed both the archive and the public service if you shoot with generous overlap and resolution and keep the originals. The discipline is to decide the split early, not after processing.

• Archive and conservation: maximum geometry, metadata, clean surfaces.
• Visitor service: small file weight, good texture, points of interest and narrative.
• One good capture can feed both — if you keep the source files and metadata.

Choosing your capture technology

Photogrammetry reconstructs a model from overlapping photographs. It is flexible, captures colour and texture, and is genuinely open to small institutions: published low-budget protocols show that small objects can be documented at publishable quality with a basic camera, controlled light and open software. For larger buildings and landscapes it also scales, especially when combined with drone capture.

Laser scanning and mobile LiDAR measure geometry quickly and accurately, which suits buildings, rooms and large monuments, but they capture little or no colour on their own, so they are usually paired with photography. Structured-light methods handle intricate small-object geometry well, but glossy, reflective and translucent surfaces — varnish, glass, polished metal — remain the hardest cases for every technique and need deliberate staging.

For built heritage you often combine terrestrial laser scanning with photogrammetry from a drone to get both precise geometry and full texture. If your budget is tight, practise on a few representative objects with photogrammetry first; that experience tells you honestly whether laser scanning is justified for the rest of the collection.

Quality, standards, and metadata that travel

Do not produce a model in a vacuum. The European VIGIE study 2020/654 on quality in 3D digitisation of tangible cultural heritage, published in April 2022, catalogued the formats, standards and methodologies in use and what determines project quality. It was written in response to the EU recommendation 2021/1970, which calls on member states to digitise monuments and sites at risk in 3D by 2030. Simplified quick-start guidance based on that study is now available to help smaller institutions adopt it without a research team.

For international consistency, treat metadata as part of the model, not an afterthought: object identifiers, author, technique, ownership and reuse conditions. Structured metadata is what makes a model machine-readable and therefore searchable, reusable in virtual exhibitions and safe to share under a licence you control.

The ICOMOS-supported study frames digital recording as essential both to preserve the past and to educate and communicate the value of tangible objects to society — a reminder that a well-documented model earns its cost twice, once as an archive and once as a public resource.

A repeatable pipeline: from frames to published model

A repeatable process matters more than expensive hardware. During capture control frame overlap (commonly 60–80% or more), steady light, absence of glare, scale and calibration. Open step-by-step protocols for small-object photography and high-resolution photogrammetry walk you through the entire route, from camera setup to post-processing of the images.

Software processing follows standard stages: image orientation, dense point cloud, polygonal mesh, then texture mapping. After reconstruction you clean noise and artefacts and simplify the mesh for its target. Control points are about not losing fine detail during simplification and not leaving holes on difficult surfaces.

Always produce two outputs: an archival master at full quality and a derived light copy for the web and apps. Store the master in broadly supported formats next to the source photographs and metadata; optimise the derived copy in polygon count and textures so it opens quickly on a visitor's phone.

• Capture: overlap, light, scale, no glare.
• Process: orientation → point cloud → mesh → texture → cleanup.
• Two outputs: an archival master and a light web/AR copy.
• Record metadata at every step, not at the end.

Turning the model into a visitor service

Augmented reality in the gallery is one of the most direct ways to make a scan useful. Visitors point a phone camera at a marked object and receive text, audio, layer comparisons or a before-restoration view. Projects in this area, such as museum AR guides, show that a handful of digitized highlights can serve both local audiences and tourists and keep working after the gallery closes, since people return to the content at home and share it.

For your own website you need an embeddable 3D viewer plus a file on your server. Practice in the heritage sector suggests models of roughly 15–40 megabytes give comfortable viewing on both mobile and desktop; larger files slow the page and discourage use. You can publish through established hosting platforms that already offer viewers and community features, or self-host and embed a viewer plugin — self-hosting gives more control but raises traffic and maintenance costs, so it fits institutions with strong IT support or only a handful of showcase models.

Hosting platforms typically let you set a licence such as Creative Commons, which matters if you want wide reuse while keeping attribution and non-commercial limits. Whichever route you choose, remember the 3D is the vehicle, not the story: add annotations, points of interest and before/after layers, or the model stays pretty but useless.

• Gallery AR: phone points at the object for text, audio and layers.
• Website: embeddable viewer plus light copies of ~15–40 MB.
• Hosting platforms with viewers and licensing, or self-hosted plugins.
• Add narrative and points of interest — the model is not the story itself.

Budget, skills, and measuring real value

Start with a pilot of a dozen significant objects rather than a full collection sweep. Bring in a contractor with laser scanning only for complex or high-value pieces; capture routine objects in-house using open protocols. That balances cost and gives your team transferable skills instead of a single outsourced dataset.

Keep the two missions separate. Archiving is long, planned and aims at coverage of the whole collection. A visitor service is short, public and built on a curated selection of objects that carry the story. Trying to solve both at once is the main cause of budget overruns.

Judge success by behaviour, not gigabytes: how many visitors opened the model or used the AR guide, how long they stayed with the object, whether they returned to view it from home. Those figures tell you which objects to digitize next and where the service is genuinely wanted rather than fashionable. Measuring use is also how you justify the next budget round to leadership and funders.