This lesson introduces the principal digital methods used to document and analyse cultural heritage sites and objects. It combines a short theoretical overview with a practical introduction to photogrammetry using Agisoft Metashape.
In the first part, we explore the main technologies currently used in heritage documentation, including:
In the second part, we focus on the workflow of Agisoft Metashape, a widely used photogrammetry software. You will learn how to generate a 3D model from photographs taken with a camera or smartphone, including the processes of image capture, photo alignment, point cloud generation, mesh creation, texturing, and tiled model creation.
No prior technical experience is required. The lesson is designed as a step-by-step introduction for beginners interested in applying digital methods to cultural heritage research and practice.
Before beginning, considering the following questions:
Digital technologies are commonly used in cultural heritage for two broad purposes:
Digital recording methods provide accurate and detailed documentation of heritage sites, monuments, and artefacts. Such records support conservation planning, condition assessment, and long-term preservation.
Digital documentation is particularly valuable when heritage is threatened by natural disasters, climate change, conflict, or urban development. Following the fire at the Notre-Dame Cathedral, for example, previous laser-scanning surveys played an important role in informing restoration efforts.
Digital heritage data can also be used to improve public access and understanding through:
Case Study: [UNESCO’s discussion of protecting heritage sites in Lebanon demonstrates how digital technologies can assist in documenting and safeguarding cultural heritage during periods of conflict] (https://www.youtube.com/watch?v=jqAAi5QFyQQ). This lesson focuses primarily on the first area—digital tools for research, documentation, and conservation—while briefly introducing applications for public engagement.
Several technologies are commonly used to capture and record heritage assets:
Before we go into details, it is useful to think of heritage digitisation as a sequence of interconnected stages rather than a single technical process.
It begins with data collection, where digital tools such as photogrammetry, laser scanning, or archival digitisation are used to capture physical, visual, or spatial information from heritage sites and objects. This is followed by data processing, which involves cleaning, aligning, scaling, and transforming raw data into usable forms such as point clouds, models, or maps. However, data do not become meaningful on their own. Through data interpretation and communication, digital outputs are analysed, contextualised, and translated into drawings, narratives, visualisations, or arguments that support research, conservation decisions, and public understanding. Finally, data sharing ensures that digital heritage data can circulate beyond a single project, through archives, repositories, publications, or open platforms, raising important questions about accessibility, standards, ethics, and long-term preservation.
Digital heritage workflows are rarely linear, which often require researchers to revisit earlier stages and refine their data collection or processing methods.
What it is Structure form Motion?
Structure from Motion (SfM) is a computer vision technique that reconstructs three-dimensional geometry from a series of overlapping two-dimensional photographs. In simple terms, it turns a collection of images into a 3D model.
SfM works by identifying matching features across multiple photographs and using these correspondences to estimate both the position of the camera and the three-dimensional structure of the scene. The resulting output is typically a point cloud, which can then be developed into a textured 3D model.
Although SfM is often grouped under the broader category of photogrammetry, there is a subtle distinction between the two. Traditional photogrammetry usually relies on known camera positions, pre-calibrated cameras, and carefully planned image acquisition. SfM, by contrast, automatically calculates camera positions and orientations from the photographs themselves. In contemporary practice, however, the terms SfM and photogrammetry are often used interchangeably.
Structure from Motion concept (http:// www. theia-sfm. org/ sfm. html)
1. Feature Detection and Matching Distinct visual features are automatically identified and matched across multiple photographs.
2. Camera Calibration and Pose Estimation The software estimates camera positions, orientations, and internal camera parameters.
3. 3D Reconstruction Matched features are used to generate a sparse point cloud, which can be further processed into dense point clouds, meshes, and textured models.
One of the most influential algorithms used for feature matching is SIFT (Scale-Invariant Feature Transform), developed by David Lowe in 1999. SIFT enables reliable feature detection across changes in scale, viewpoint, and lighting, making it particularly effective for photogrammetric reconstruction.
For readers interested in the theoretical foundations of SfM, see: Ullman, Shimon, ‘The Interpretation of Structure from Motion’, Proceedings of the Royal Society of London. Series B. Biological Sciences, 203.1153 (1979), 405–26 https://doi.org/10.1098/rspb.1979.0006
Lowe, David G., ‘Object Recognition from Local Scale-Invariant Features’, in Proceedings of the Seventh IEEE International Conference on Computer Vision, 2 vols (Los Alamitos, CA: IEEE Computer Society Press, 1999), II, 1150–57 https://doi.org/10.1109/ICCV.1999.790410
Early applications of the SIFT algorithm:
A landmark application of SfM was the Photo Tourism project developed by Noah Snavely, Steven Seitz, and Richard Szeliski in 2006. The project demonstrated how large collections of publicly available photographs could be used to reconstruct famous landmarks in three dimensions, laying the foundation for many contemporary photogrammetry workflows.
Photo Tourism, see Snavely, Noah, Steven M. Seitz, and Richard Szeliski, ‘Photo Tourism: Exploring Photo Collections in 3D’, ACM Transactions on Graphics (Proceedings of SIGGRAPH 2006), 25.3 (2006), 835–46 https://doi.org/10.1145/1141911.1141964
Several software packages implement SfM and multi-view stereo (MVS) workflows:
SfM has become one of the most widely adopted methods in cultural heritage documentation because it offers several advantages:
Laser scanning is a surveying technique that uses laser light to measure the position of points on an object’s surface and record them as three-dimensional coordinates. Unlike photogrammetry, which reconstructs geometry from photographs, laser scanning directly measures distance and spatial position.
Because millions of measurements can be captured in a short period of time, laser scanners are capable of producing highly accurate and detailed records of buildings, monuments, archaeological sites, and landscapes.
Laser scanners emit pulses of light and measure how the light returns after striking a surface. The scanner then calculates the position of each point in three-dimensional space.
Depending on the instrument, laser systems may use:
Laser scanning systems generally fall into two categories:
1. Range-Based Scanners (Time-of-Flight or Phase Shift) These scanners calculate distance by measuring how long it takes a laser pulse to travel to an object and return to the sensor. They are commonly used for surveying buildings, archaeological sites, and large landscapes.
2. Triangulation Scanners These scanners determine position through geometric triangulation between the laser source, the sensor, and the target object. They typically achieve higher precision but are designed for smaller objects and shorter scanning distances. Different ranges of static laser scanning
After fieldwork, scan data must be processed and combined into a complete model.
Common tasks include:
Common software packages include:
Both photogrammetry and laser scanning produce three-dimensional data, but they collect information in different ways.
| Photogrammetry (SfM) | Laser Scanning (TLS) |
|---|---|
| Passive technique | Active technique |
| Uses photographs | Uses laser measurements |
| Reconstructs geometry from image matching | Directly measures distance |
| Low-cost and accessible | Requires specialised equipment |
| Produces high-quality textures and colour information | Produces highly accurate geometry |
| Processing is largely automated | Registration and processing often require additional user input |
| Suitable for objects, buildings, and landscapes | Particularly effective for complex architecture and large sites |
In practice, many heritage projects combine both methods. Laser scanning provides highly accurate geometry, while photogrammetry contributes detailed colour and texture information.
A point cloud is the primary output of both photogrammetry and laser scanning. It consists of a collection of points in a three-dimensional coordinate system, defined by X, Y, and Z coordinates. Together, these points describe the surface geometry of an object or site.
Although both techniques generate point clouds, the nature of the data differs:
| Photogrammetry | Laser Scanning |
|---|---|
| Point cloud generated after image processing | Point cloud generated directly during scanning |
| Usually unstructured | Usually structured and systematically acquired |
| Requires scaling and georeferencing | Recorded at true scale (1:1) |
| Often contains colour information derived from photographs | May contain colour, reflectance values, surface normals, etc. |
Simultaneous Localisation and Mapping (SLAM) is a technology that allows a device to determine its own position while simultaneously creating a map of its surroundings.
In heritage documentation, SLAM is commonly combined with LiDAR sensors to create mobile mapping systems that capture three-dimensional data as the operator walks through a site. Unlike traditional laser scanning, which requires the scanner to be placed at multiple fixed positions, SLAM systems continuously collect data while moving.
This makes SLAM particularly useful for:
A SLAM system continuously combines information from multiple sensors, including:
As the operator moves through the environment, the system identifies overlapping features and continuously updates both its location and the emerging 3D map.
GeoSLAM ZEB Horizon
Leica BLK 2GO
Structured-light scanning is a high-accuracy, non-contact method for capturing three-dimensional geometry. Instead of measuring distance with laser pulses, the scanner projects a known pattern of light—typically stripes or grids—onto an object and records how the pattern deforms across the surface.
By analysing these distortions, the system reconstructs a highly detailed 3D model.
Structured-light scanners are commonly used for:
The process involves three steps:
Because the projected pattern is precisely controlled, structured-light scanning can achieve very high levels of accuracy and detail.
Example: FARO Freestyle 2
The FARO Freestyle 2 is a handheld structured-light scanner designed for capturing complex surfaces and confined spaces.
Key features include:
Recent advances in mobile technology have made 3D documentation more accessible than ever. Several smartphone applications can generate three-dimensional models using photogrammetry or LiDAR sensors (available on some devices).
While mobile applications cannot generally match the accuracy and reliability of professional photogrammetry or laser-scanning workflows, they provide a quick and accessible way to create preliminary records of buildings, objects, and archaeological features.
Mobile scanning applications are particularly useful for:
RealityScan is a mobile photogrammetry application developed by Epic Games. Users capture a series of overlapping photographs, which are automatically processed into a textured 3D model.
Key features include:
Polycam allows users to create 3D models using either photogrammetry or LiDAR (on supported devices).
Key features include:
One of the most exciting applications of heritage digitisation is the ability to transform a digital model into a physical object through 3D printing.
Once a heritage asset has been documented using photogrammetry, laser scanning, or other surveying methods, the resulting digital model can be used to create accurate physical replicas at full scale or reduced scale. This process connects the digital and physical worlds, allowing heritage objects to be studied, conserved, and exhibited in new ways.
Research and Education Three-dimensional replicas allow researchers, students, and visitors to handle and examine objects that may otherwise be inaccessible due to their fragility, rarity, or location.
Conservation and Restoration Digital models can be used to recreate missing or damaged elements of heritage objects and buildings. Replicas may be employed for testing conservation interventions or, in some cases, for the reconstruction of lost components.
Museums and Exhibitions 3D-printed replicas enable museums to display accurate copies of important artefacts while reducing risks to the originals. Replicas can also improve accessibility by allowing visitors to touch and interact with objects that would normally be protected behind glass.
A notable example is the Hellenistic bronze sculpture Boxer at Rest, housed in National Roman Museum. For an exhibition in Beijing in 2022, a high-quality replica was produced using digital surveying and advanced fabrication techniques, allowing the work to be displayed internationally while preserving the original sculpture.
The project demonstrates how digital heritage workflows can extend beyond documentation to support cultural exchange, conservation, and public engagement.
Reproducing Art is ‘Art’: The Boxer and Its Replica (video).
Virtual, Augmented, and Mixed Reality (VR/AR/MR) technologies are increasingly used in cultural heritage to create immersive and interactive experiences of sites, objects, and historical environments.
These technologies allow users to move beyond static documentation and engage with heritage in embodied and spatial ways.
Consider the following questions:
How faithfully do digital reconstructions represent historical reality? What is the status of a virtual reconstruction in relation to the original heritage object or site? How are historical uncertainties visualised or simplified in immersive environments?
Digital heritage work is most effective when guided by clear research and communication goals. The choice of tools should always follow from the question being asked, rather than the other way around.
A successful workflow typically depends on:
‘If you fail to plan, you are planning to fail.’
In digital heritage practice, planning is not only technical but also conceptual. It involves selecting appropriate methods, understanding their limitations, and combining them into an integrated approach that aligns with the research question.
Software used: Agisoft Metashape
Download from the official website: https://www.agisoft.com/downloads/installer/
Choose the Professional Edition. A 30-day free trial is available. After the trial period, the software can still be used in Demo mode, or unlocked with a paid license.
The basic photogrammetry workflow in Metashape is as follows:
In the following sections, we will go through this process step by step to create a 3D model from your own images.
Choose an object or a space you would like to reconstruct in 3D.
Not all subjects are suitable for photogrammetry. Avoid objects that are:
Photographs for Agisoft Metashape can be taken with any digital camera (including smartphones), as long as consistent capture guidelines are followed. Image quality and overlap are essential for successful reconstruction.
Whatever camera you use, follow these principles:
For successful reconstruction, ensure:
Overlap and camera coverage diagram
If the object has repetitive patterns or lacks distinct visual features, fiducial markers can be added during photography to improve feature matching and alignment.
This section walks through the full photogrammetry workflow in Agisoft Metashape, from image import to final export.
First, import your images into Metashape.
You can do this in several ways:
Select all images and click Open. A new chunk will be created containing your dataset.
Go to Workflow → Align Photos In most cases, you can keep the default settings (e.g. medium accuracy and generic preselection). After processing, Metashape generates a sparse point cloud and estimates camera positions. In the Workspace panel, you can check how many images were successfully aligned. Ideally, all photos should be included. You can navigate the scene using your mouse or trackpad (orbit, zoom, and pan). Camera positions can be displayed by enabling: Model → Show/Hide Items → Show Cameras
Before generating the dense cloud, define the working area to reduce unnecessary computation.
Go to: Region → Resize Region.
Adjust the bounding box to include only the object of interest. Ensure all important parts remain inside the region from multiple viewing angles.
Then run: Workflow → Build Dense Cloud (quality: medium, depth filtering: aggressive)
Once complete, the dense cloud should contain millions of points and clearly represent the object’s surface geometry.
The dense cloud may include unwanted points (background noise or reconstruction errors).
Use selection tools to:
This step improves mesh quality in later stages.
Remember to save your project regularly.
Go to Workflow → Build Mesh Choose settings depending on your subject:
For object:
For space/terrain:
The mesh converts the point cloud into a continuous surface.
Once the mesh is generated, you can apply image-based texture.
Go to Workflow → Build Texture
This step projects the original photographs onto the mesh, producing a photorealistic surface representation. Texture is one of the main advantages of photogrammetry compared to other 3D capture methods.
Tiled models are designed for large-scale visualisation (e.g. buildings, cities, landscapes).
Go to Workflow → Build Tiled Model
This format supports efficient streaming and rendering of high-resolution models by dividing them into hierarchical tiles.
Tiled models are typically generated from mesh, point cloud, or depth maps and are automatically textured from source images.
When satisfied with the result, export your model via:
File → Export → Export Model
You can export different outputs depending on your needs:
After export, you will typically obtain three files:
These files must remain in the same folder for correct display. Some software will display the image texture automatically when you open the OBJ and in other software you might need to re-assign the image texture to the material.
You can view exported models using:
To share models online:
For guidance on publishing to Sketchfab, see:
http://cycollection.com/visualthinking/links.html
Written by Jiayao Jiang, jj596@cam.ac.uk, 2026-08-18
Licence: CC BY-SA 4.0
