3D digitizing archaeological finds
Hi! My name is Annukka Debenjak-Ijäs. I started to work with 3D digitization as an archaeology student six years ago. As a field archaeologist, I have always had a strong interest in different technologies, which could ease the archaeological documentation process. After some experiences with field documentation using photogrammetry, a shift to smaller archaeological objects was an interesting change. =)
Currently, I study the 3D digitization process of archaeological finds in a one-year joint research project of the Department of Archaeology at the University of Turku, the Museum Centre of Turku, and the Aboa Vetus & Ars Nova Museum. The emphasis is on the analysis of the overall procedure from selecting and modelling of the objects to storing, distributing, and using the models in academic research, education, and public engagement like museum exhibitions. The project is funded by the Ministry of Education and Culture in Finland.
In my work I see the many benefits of 3D digitizing selected archaeological finds. One of the find groups I digitized during this project is an unpublished collection of Iron Age bronze fibulae, which are fragile and accessible only in Turku at the Department of Archaeology. Now the 3D digitized fibulae are free to use also for researchers abroad. International research and sharing of the find materials is important especially in Finland, where the influences of eastern and western cultures are always visible in the find material.
The corona crisis affecting our society in the spring of 2020 highlighted the importance of 3D material in archaeological education—compared to two-dimensional images, the 3D digitizations provide a better way to learn to recognize different find types. Our test with a ceramic sherd from the late Stone Age, with added annotations from a ceramic expert, inspired a separate 3D digitization project focused on an educational collection of Finnish prehistoric ceramic types.
The Turku Museum Centre
The Turku Museum Centre consists of the local museums owned by the City of Turku: The Turku Castle, the WAM Turku City Art Museum, the Pharmacy Museum and the Qwensel House, the Luostarinmäki Outdoor Museum, the Biological Museum, and the Kylämäki Village of Living History in Kurala. The Museum Centre functions also as the Provincial Museum of Southwest Finland, taking care of a diverse collection of archaeological finds, including around 400,000 items from Prehistory to the Post-Medieval period.
As the Museum Centre has a strong interest in developing the use of digital technologies in the Finnish cultural heritage field, it invested in two high-class hand-held 3D scanners, Artec Leo and Artec Space Spider. Photogrammetric 3D digitization is carried out with the RealityCapture software. In addition to the 3D digitization of archaeological finds, other samples from the museum collections are 3D digitized by the museum staff. 3D models of samples of the Biological Museum, Medieval toys, toy animals, and, e.g., a collection of pipes are published in the Sketchfab collections of the Museum Centre.
How do I do it?
At the start, small-object photogrammetry seemed quite hard because of the short imaging distances which resulted in a very short depth of field and a blurrily textured 3D model. Thus, the first task was to solve the DoF challenges. From the different options, such as using a tilt-shift lens, focus stacking, and masking the unsharp parts of the images, I chose the last one. After some trials, errors and improvements, my 3D digitizing workflow looks as follows:
Choosing between photogrammetry and the Artec Space Spider
Easily scannable finds, with a simple shape, matte surface, and a size suitable for the Artec 3D scanners were scanned following the 3D Scanning Guidelines for Skeletal Remains with Artec Studio 11 established by the University of Sheffield (Campanacho 2017).
Finds with a shiny surface, complex geometry or small size, such as this 3 cm high lead figurine, were digitized using photogrammetry.
Equipment and imaging workflow
My photogrammetric workflow follows mainly the instructions described in Closing the seams: resolving frequently encountered issues in photogrammetric modelling by Matthew Magnani et al. (2016). I use a Nikon D750 with a Nikkor 60 mm macro lens. For finds with a shiny surface I add a circular polarizing photographic filter. With a prime lens, it is possible to gain more accurate results than with a zoom lens. The finds are placed in a Foldio 360 light tent on a turntable and lighted with a diffuse 5500 Kelvin light. As the imaging equipment has to be moved between the museums and the university, this light solution fits our needs well.
The Foldio turntable and camera are controlled by a mobile phone app, which remotely turns the turntable and shoots 24-48 image circles of the find. After taking 2-4 image circles from the first position of the find, the find is turned around and photographed from the next side. For simply shaped finds, 2 positions are enough, but complex finds have to be photographed in 3-4 different positions.
To set the 3D digitization to the right scale, a small scale bar is set next to the find when photographing the first position. Nowadays RealityCapture also supports coded markers, so the next step for me is to add markers with known positions on the turntable.
To ensure a wide depth of field and sharp image quality, I use a relatively small aperture size of f/16, and set the ISO value to 100. This leads to long shutter speeds, so the camera is placed on a tripod to prevent motion blur. I shoot the images in RAW format and convert them to .tif using RAWtherapee. During this workflow, the brightness and white balance of the images can be adjusted, which is necessary when using the polarizing photographic filter.
Processing with RealityCapture, Metashape, and Blender
I process the images in a photogrammetric software called RealityCapture. After aligning the images, the mesh is processed with high detail settings. Before texturing the mesh, I switch to masked images produced in Agisoft Metashape by changing the image format from .tif to .png. This workflow results in more sharp textures, because the blurry parts are excluded from the images. I noticed that it’s possible to speed up the masking process by keeping it in mind during the imaging process. I aim to take the images so that the sharp and unsharp parts of the image are always at the same place in one image circle. I then produce only one mask for each image circle and copy it to the other images.
The textured mesh is exported from RealityCapture as an OBJ file and imported to Blender for post-processing. This includes setting the 3D model to the origin point, turning it right-side-up and checking that the scale is correct. Sometimes some white colour from the images’ background appears in the texture on the edges of the find. These can be repaired with the texture painting tool. This and other changes in the texture or mesh are stored as a “before and after” image pair and mentioned in the metadata records of the 3D digitization.
To keep the information about the find size at hand if, for example, the file is imported into another program with wrong unit settings, a digital scale bar is added next to the 3D digitization before exporting the file as OBJ and glTF_embedded. For easy visualisation in Sketchfab and other online platforms, I also generate a simplified mesh of the find in Blender using the decimating tool and export it as glTF_embedded.
Dissemination and storing of the 3D digitizations
The long-term storing of 3D files is still a challenge to be faced by cultural heritage institutions. Although there is no perfect solution, it’s still important to do our best to store both the digitizations and the necessary information about the processing of the digitization. This allows future users to estimate the usability of the digitizations for their needs.
In our project, the files are stored on the servers of each partner organisation and shared online via Sketchfab and the webpages of the partner organisations. Files resulting from the digitization process are stored together with the original OBJ and glTF files and the simplified glTF file. These files include the original images in .tif and .dng file formats, the masked .png images, the RealityCapture project file, the images of texture or mesh post-processing and the process description and file information as a .pdf file.
A great amount of the value of 3D digitizations comes from the ability to share them online across the world. For this purpose, Sketchfab is an easy-to-use platform. In my opinion, the fragmentation of information on numerous small webpages is a real problem, so I prefer publishing the digitizations on a well known and wide-reaching community. In the future, we plan to share our material also in Europeana via the Share3D application, and the national cultural heritage platform Finna.
By providing the opportunity to download and reuse the 3D digitizations produced in our project, we encourage people to interact with cultural heritage objects. I think this also lowers the barrier to engage with heritage that is “locked away” in Museums and other institutions.
Tips and thoughts about the future
Open your eyes—my mistake was to first try things out without searching for a lot of information, and then learning that someone had already solved the problems I faced. Just browse around Sketchfab, Google or other sites to see if someone else is working with the same kind of material or research questions as you, and don’t be afraid to contact them. I have mailed numerous people and asked the “stupid questions” that a beginner asks, and gotten wonderful advice, which helped me progress. =)
How do I see 3D changing the way heritage organizations work in the future?
I think 3D digitizations should be treated more like a tool—not a destination—for archaeological research. This includes standards in the project planning, metadata storing and sharing of 3D material. Organisations such as the Smithsonian Digitization Program, Europeana, and GLAM3D are working hard to establish and disseminate best practices regarding the whole life cycle of a 3D digitization.
In the future, I hope to see more diversity in the production of 3D content of cultural heritage. At least now, usually only a part of museum collections are 3D digitized. Including the local community in the decisions of what should be digitized, as well as how and why, is a way to share experiences and values about cultural heritage that otherwise might be forgotten. Via digital 3D content we can unite layers of knowledge to physical objects, and share them to a wider audience. I think this will lead to a more egalitarian, interactive way of handling cultural heritage in heritage organizations.
My Sketchfab favourite
Browsing Sketchfab for different cultural heritage digitizations is a good way to learn about different methodological approaches. Recently I saw these impressively sharp 3D digitizations of archaeological finds at the Roman legionary fortress of Carnuntum in Austria made by Harald at noe-3d.at. After contacting Harald I learned about the focus stacking software Helicon focus, which will be my next area to study deeper.