Why scan an object at all
A 3D scan records an object’s shape, and often its surface color, as data that can be turned, zoomed and examined on a screen. For exhibits, that can mean letting visitors handle a fragile piece virtually, showing the back or underside of an object that sits in a case, or giving researchers measurements without disturbing the original. A scan does not replace the object, and it is worth deciding early what question the scan is meant to answer.
Common scanning approaches
Most projects use one of a few approaches. Each suits different sizes and materials, and the best choice often depends on the object more than on the budget.
Photogrammetry
Photogrammetry builds a 3D model from many overlapping photographs taken from different angles. Software finds matching features across the images and calculates the object’s shape. It needs relatively modest equipment, since a camera, steady lighting and a turntable or careful movement around the object are often enough, and it tends to produce good surface color. It depends on the object having visible texture and on the photographs being sharp, evenly lit and well overlapped. Results can vary a lot with technique.
Structured light scanning
A structured light scanner projects a pattern, often stripes or a grid, onto the object and reads how the pattern bends over its surface. It is commonly used for small and medium objects and can capture fine geometric detail. Many handheld and tabletop devices work this way. Performance can drop in bright ambient light, and very dark or glossy surfaces may give gaps.
Laser scanning
Laser scanners measure distance by sweeping a laser across the object. Close-range versions can record very fine surface detail, and longer-range versions are used for large items such as architecture, vehicles or outdoor sculpture. They are usually more specialized and costly than photogrammetry, and often come through a service provider or a partner institution.
Computed tomography
Medical-style or industrial CT scanning records the inside of an object as well as its outside. It is used for things like mummies, sealed vessels or composite objects. It is a specialist process, and the output is a volume of data, not a simple surface model.
What you get back: common outputs
Scanning produces more than one kind of file, and knowing the pieces helps in planning storage and delivery.
- Point cloud. A large set of measured points in space, sometimes with color. It is a raw or early form of the data.
- Mesh. A surface made of connected triangles. This is the form most viewers and game-style engines use. Its detail depends on how many triangles it has.
- Texture maps. Image files wrapped onto the mesh to carry color and fine surface appearance.
- Raw captures. The original photographs or scanner data, which are worth keeping so the model can be rebuilt later with better software.
Several file formats exist for meshes and textures. Some are simple and widely supported, some bundle the mesh and textures together for web viewing, and some are intended for archives. A project usually keeps a high-detail master for preservation and creates lighter derivatives for screens, since a master mesh can be far too heavy to run smoothly on a gallery computer or phone. If the models will appear on the web, the practical limits are covered under 3D viewers.
Documenting the scan matters as much as the model. Record the method, equipment type, date, the person responsible and any processing steps. Those details are part of the metadata a collection needs, and they help the next person judge how far the model can be trusted.
What scanning captures poorly
Every method has blind spots. Knowing them helps set expectations before anyone is hired or equipment is borrowed.
- Shiny and mirror-like surfaces. Polished metal, glazes and varnish reflect light unpredictably, which confuses both photogrammetry and structured light.
- Transparent and translucent materials. Glass, some plastics, crystal and thin porcelain often scan as holes, distortions or noise.
- Very dark or black surfaces. These absorb light and can return little usable data.
- Thin, fine or hair-like parts. Fringe, feathers, filaments, delicate wire and lace are hard to separate from the background and often come out as blobs.
- Deep cavities and undercuts. If the scanner or camera cannot see into a recess, the model will have a hole or a guess there.
- Soft, moving or deformable items. Textiles that shift, or any object that cannot be kept still, can misalign the data.
- Featureless surfaces. For photogrammetry, a uniform smooth surface gives the software nothing to match.
Workarounds exist, such as temporary dulling sprays or powders, but for collection objects these are often not acceptable because they could change or harm the surface. Conservation staff should make that call. In many cases the honest answer is to scan what scans well, or to choose a photographic approach instead, as described for object photography.
Scale, accuracy and honesty
A model that looks convincing is not automatically accurate. Check dimensions against known measurements, and include a scale reference during capture. If holes were filled or surfaces smoothed in processing, say so in the documentation. Where a scan is presented to visitors as a record of the object, it helps to be open about what has been cleaned up or reconstructed.
Planning for display
Think about how visitors will use the model. A turntable view with a few preset angles is far lighter and easier for casual visitors than a free-rotating model, and it works more reliably on touchscreens. Free rotation suits specialist users who want to inspect details. Text descriptions and alternative views should accompany any model, since a 3D viewer may be hard to use for some visitors, a point raised in the discussion of screen readers and tactile models. A printed replica from a scan is one way to offer a touchable version, which is the subject of tactile models.
Finally, consider rights and cultural context. Some objects carry restrictions on how they can be shown or shared, and consulting source communities before scanning and publishing is good practice.