Archiving the Third Dimension: Production, Maintenance and Use of 3D Models in Cultural Heritage Management

Abstract

In 2017, the Archaeology Department of the Canton Zurich defined photogrammetry in Agisoft Metashape as a new standard. The years have shown the advantages of the method, especially when documenting complex 3D structures, like skeletons or ruins, on site. While generating 3D models was a well-trodden path, the storage of 3D models needed some exploration. The first challenge was to incorporate 3D data into a pre-existing archive system. The second challenge was to secure both easy access for everyone and long-term storage at the same time. We propose a solution where access and storage are treated as two separate issues. This allows the content to be viewed without technological restrictions (e.g. as 3D PDF or VR application). While independently the same models are stored in a normalised geometry file (e.g. OBJ file) for long term use. Since long-term standards for 3D data have yet to be established, we have decided to provide a temporary backup system by keeping the raw data (photos) together with the processing report for replication purposes. Our solution comes with a lot of redundancy. However, this is still a trial and error approach, one that we would like to work on with fellow 3D enthusiasts.

1 Introduction

The value of photogrammetry for documenting archaeological features is well proven (cf. De Reu et al. 2014). But has the technology been fully established within archaeological procedures regarding rescue excvations? What criteria must a technology meet in order to be described as established? Since archaeology is a science in which research is seldom based on the object itself but almost always on the primary documentation, ensuring the preservation of that primary documentation is intrinsically crucial. So long as this cannot be guaranteed in the case of image-based modelling, the technology, strictly speaking, should only be used, at most, as a supplementary method of documentation.

As a state enterprise, the Archaeology Department of the Canton Zurich is tasked with locating areas of potential archaeological importance and, if necessary, carrying out rescue excavations. The documentation is kept and made available for future research. The results of excavations are also made accessible to the public.

Since 2017, the Department has been using photogrammetry as a standard method of documentation. The advantages of this method were apparent right from the start. However, staff were not properly trained yet, there were no standards on which to base procedures, and, above all, an inadequate archival infrastructure meant that results could not be properly archived. These problems prevented the technology from becoming fully established, and so a project was initiated, tasked with finding solutions. The resulting concept envisages being able to archive data from 2021 onwards. The archive will then be kept under regular review until 2027, in the hope that by the end of the project, we will be in a position to propose a long-term archiving solution. Section 12.2 of this article examines the areas of application of image-based modelling and attempts to infer from these the future requirements of archive users. In Section 12.3 these requirements are integrated into an archive system.

Note: Currently, the project—like this paper—is addressing only data relating to excavation documentation, not to documentation of findings. The project will also address the archiving of laser-scan data, but that is beyond the remit of this paper.

2 Areas of Application of Photogrammetry

Traditionally, the documentation of archaeological excavations by our Department rests on three pillars:

• 1. Photographs, which give an objective impression of the situation being documented.

• 2. Drawings, which precisely localise and interpret the features (cf. Morgan and Wright 2018).

• 3. Journals, which describe the features in words.

The documentation is used in archaeological research and made available to the public in an appropriate form.

3D recording plays a role in almost all areas of our work and brings with it archiving requirements, as described below.

2.1 Objective Primary Documentation

Basically, 3D recording gives an almost objective spatial impression, which reflects, without commentary, the situation on site. In this way, it is similar to traditional photography (cf. Fouriaux, Introduction of Chap. 5 in this book). Photogrammetry really comes into its own where photography comes up against its limitations: when features present strongly three-dimensional characteristics and therefore cannot be recorded in a single photograph. In the case of walls, kilns, wells, baths—in short, any structures with strongly three-dimensional characteristics, photogrammetry brings added value. Because of the additional information it includes, its use should be imperative in the documentation of such features, in so far as the necessary resources are available. On the other hand, an simple, flat archaeological layer can be recorded without difficulty by a photograph. When the image has been rectified, it affords almost exactly as much information as photogrammetry. In this case, 3D recording does not need to be used.

3D models make it possible to examine features from a single image and gain an impression of their totality, instead of having to mentally piece together separate images to make up the whole (Fig. 12.1). As well as strongly three-dimensional features, 3D images can thus also be used to capture complex, cross-trench stratigraphic sequences.

Fig. 12.1

Example of a complex structure, a 3D model of an oven. Excavation of a medieval monastery in Winterthur, Switzerland. (Image: Kantonsarchäologie Zürich)

Burials are a particular case in point. Skeletons themselves present a high degree of three-dimensionality and are hard to record satisfactorily with photographs. If there are grave goods as well, the situation becomes even more complex. Burials rich in grave goods are routinely recorded three-dimensionally; those without grave goods may be too, depending on whether anthropological questions require a more detailed view of the skeleton.

The need for an archive to offer objective documentation is self-evident. Future researchers must be able to extract new knowledge from archived material. In the case of photographs, archive users should be able to call up and examine very high-resolution images via a database. Similar requirements exist in the case of photogrammetry. Users want to be able to make full use of the images, in other words, to examine them in three-dimensional space and in sufficiently high resolution, without the need for pre-existing technical expertise in working with 3D models. Since the 3D model contains geographical information, they also need to be able to take measurements or retrieve coordinates.

This calls for a viewer with an easy-to-use, device-independent navigation and orientation function.

2.2 Localisation and Interpretation

The products of 3D recording are also used as the basis for the second pillar of excavation documentation mentioned above: site drawings. These traditionally have two purposes:

• 1. Localising the feature in its surroundings

• 2. Interpreting the feature

Traditionally, drawings are made in the field on the basis of a simple survey grid. Measuring every stone by hand is very laborious. In the case of complex objects, such as courses of brick or masonry, drawings are increasingly being made from orthophotos, which, in turn, are computed from photogrammetric images (cf. Brünenberg et al., Chap. 4 in this book). There may be other reasons, apart from time pressure, for employing this procedure; for instance, if the feature concerned is located at a great height or at the bottom of an inaccessible trench and therefore poses safety concerns (Fig. 12.2). Here, too, an orthophoto can provide the basis for the excavation drawing. The orthophoto itself is part of the archaeological documentation and must therefore be preserved in the interests of providing a complete trail of scientific evidence. The most important aspect of the orthophoto, apart from the actual image data, is the associated geoinformation, and this, too, must be available in the archive.

Fig. 12.2

Example for a structure with physically inaccessible features. 3D-model of the medieval fortification ‘Erdmannliloch’ in the Canton of Zurich, Switzerland. The documentation was made due to slow destruction of the findings by use of hikers. (Image: Kantonsarchäologie Zürich)

2.3 Archaeological Research

Our Department still has relatively little experience of using 3D models in scientific research. Up until now, the visual aspect of image-based modelling was the main contribution to understanding certain findings. For this usage, the archival requirements have already been discussed in Sect. 12.2.1. 3D models can also, however, assume an active role in archaeological research.

Unfortunately, there has not yet been any comprehensive compilation of analytical methods which rely on 3D recording. Our attempt, shown in Table 12.1, therefore makes no claim to be complete.

Table 12.1 Exemplary compilation of analysis methods based on 3D recording

Since the applications are extremely varied, it is difficult to tell, when archiving data for future use, in what format or resolution they may be needed. Moreover, new research methods will probably be developed in the future. Ours is therefore merely an initial, tentative approach to the problem:

The types of analysis listed, all use the generated mesh. On the other hand, texture is not relevant for some of the applications; it is much more important that the mesh offers high resolution and high edge definition in the area of the relevant features. If the 3D model is to be used as the basis for a reconstruction, however, texture is often needed, while high resolution is somewhat less important.

When creating and archiving 3D images, therefore, it makes sense to try and think of every probable future research question at the outset. However, since it is seldom possible to anticipate every future area of enquiry at the moment when the photograph is taken, care should be taken to ensure that a high-resolution mesh can be exported from the archive. At the same time, texture should not be neglected, since it may be needed in the case of a reconstruction. Moreover, it should be possible to open the data in widely used modelling programs ranging from ArcGIS Pro to Cinema 4D; in other words, data should be stored in as widely-used a format as possible.

2.4 Conveying Information to Specialists and the General Public

Just as difficult to envisage as the future needs of scientific research are the future methods by which information will be conveyed to both specialists and the general public. There are many different ways in which 3D recording can be used to present archaeological features and these are constantly changing.

Currently, our Department makes 3D images available to the public in two ways.

• Via platforms like Sketchfab, 3D models can be embedded into blog posts or linked to scientific publications. Blog posts are a way for our department to bring our daily work to a broader public. We did not yet use links to sketchfab in publications for scientific audience yet, but are working on the framework now.

• In the context of public outreach work or scientific conferences, we use virtual- or augmented-reality applications to display 3D recordings via smart devices or 3D glasses. We have implemented those applications in specific, temporary live events. There was only limited interaction with the 3D-models and narrative was given by live persons.

We have had an overwhelming response, particularly from the general public, to augmented-reality applications. Even people who are not tech-savvy find them easy to access and experience the vanished features almost as if they were real.

The ability to bring physically lost features back to life in this way has been clearly specified as an important requirement by the Department. For this reason, it must also be possible for archived data to be retrieved and processed for visualisation. Visualisation relies on a mesh. Texture is essential, however, since here, unlike in spatial analysis, the visual impression is what counts. In many applications, the resolution and texture of the model currently have to be greatly reduced. Since the requirements of the transporting media are constantly changing, it must be possible to export a product from the archive that is suited to the medium concerned. The archived image must therefore be available in a format whose resolution is easily scalable.

2.5 Summary of User Requirements for a 3D Data Archive

We have now described the various areas in which 3D models play a role in the documentation of archaeological excavations. For every area, we have worked out what demands would be made of the archive. To summarise (Fig. 12.3) supported by the results of our study:

• Everyday users need simple access: As objective primary documentation, 3D models must be able to be examined and measured in three dimensions by every archive user without needing to use a complex 3D software.

• Specialists need high-resolution data: for archaeological research, it must be possible to export a high-resolution mesh with texture.

• Mediation requires situation-specific data: for visualisation for educational purposes, it must be possible to export a mesh with texture in a resolution suitable for the medium being used.

• The work of the Department requires georeferencing: If orthophotos are created, these should be retrievable along with their geoinformation.

In terms of user analysis, it can be established that where the use of 3D models is concerned, there are different user groups (ordinary users vs. specialists) with different requirements for data quality and data representation. It is now a question of incorporating these different needs into the archive design.

Fig. 12.3

Simplified user requirements for a 3D data archive

3 Archiving Photogrammetry

3.1 New Challenges

The requirements summarised in Sect. 12.2.5 for the long-term archiving of 3D data demonstrate a multiplicity of sometimes conflicting demands. It is therefore clear that, unlike in the case of traditional text and image data, there is no one data format for 3D data that can cater for all requirements at once (e.g. PDF/A for texts or TIFF for images). There is therefore a need for a radical rethinking of old archive repositories—which have usually evolved from analogue structures. Instead of trying to find a tailored solution for each new problem, it would be much more sustainable to integrate scope for technological development as part of the archiving process. Thought therefore needs to be given to:

• Separating access from archiving: in order to give user requirements more space and the opportunity for development, the way data are used must be considered separately from the way they are archived.

• Archiving of datasets: pragmatic solutions are needed for storing data objects which consist of several individual files (e.g. ZIP-File, Georeferenced Images)—something which is becoming increasingly common, particularly with interactive formats.

• Future-oriented archive systems: there is a need for fundamentally simpler and at the same time more flexible archive systems, into which future data formats can easily be incorporated.

This required approach is supported by the ‘Open Archival Information System’, which we consider to be the optimal basis for archiving 3D data, as well as for other future formats.

3.2 The Open Archival Information System (OAIS)

OAIS is a reference model for long-term digital archiving. It has its origins in international space travel and was first presented by the Consultative Committee for Space Data Systems (CCSDS 2012). Since 2003, it has been listed as an international standard, ISO 14721, and has been constantly expanded (ISO 2020). As a reference model, OAIS is simply a concept for digital long-term archiving which provides for a clear division of roles, defined task areas and a shared terminology. It does not, however, specify the technical means, detailed procedures, and agents by which the concept should be implemented. It therefore offers a shared basis for digital long-term archiving which can be adopted by any archive and implemented in a way that meets everyone’s own requirements (CCSDS 2012, Chap. 1.2; applied in KOST 2020).

In our current project, the emphasis lies on the storage of the same information in multiple form, according to different application purposes. This type of procedure can be operationalised in OAIS using so-called ‘information packages’

OAIS information packages are containers for data, created for specific areas of application, for which the same sets of archiving rules apply. OAIS defines three areas of application: data ingestion, data archiving, and data access. This makes it possible to deposit the same information in three different information packages, with different data formats and different rules. The definition of these formats and rules must be established by all the participants in advance. OAIS defines three roles of participants which should be taken into account when making this decision: Producer, Management and User. With their complementary points of view, the occupants of these roles have a shared responsibility for developing a suitable infrastructure, efficient workflows, and a sound archiving strategy.

3.3 The OAIS Archiving Strategy

Based on the OAIS reference model, we are now able to present a first archiving strategy for photogrammetry in the Archaeology Department of the Canton Zurich (Fig. 12.4). It was developed through wide-ranging discussions within the Department, supplemented by specialist input from research groups (DIG 2020; Hostettler et al. 2019), specialist units (e.g. KOST 2020; Library of Congress 2019) and some feedback from universities. This broad-based exchange ensures the equal involvement of producers, managers and users right from the start.

Fig. 12.4

Archival strategy for the storage of 3D data in the Archaeology Department of Canton Zurich based on the OAIS reference model

In everyday use, the three roles are performed by mainly internal personal with the following definitions, tasks and responsibilities:

• Producers: members of the internal technical staff with specialist knowledge of 3D recording. Since image-based modelling requires specialist expertise, the producer’s role is clearly important. They are not only responsible for the correct production of data but will need to be consulted by management for years to come about such things as data output to third parties and data conversion.

• Management: internal, professionally trained archive staff tasked with long-term archiving. This includes regular checking and, if necessary, reconfiguration of data in the long-term archive to ensure that data formats remain readable and up-to-date (Life Cycle Monitoring).

• Users: internal staff members and external archive users with different professional, qualitative and technical expectations of 3D models (see Sect. 12.2). It is their responsibility to present their requirements or new needs to the management in a timely and precise manner.

When handling data, we strive for a clear separation between the areas of data ingestion, data archiving, data access and data management. Therefore, we work with a separate information package for each area. Furthermore, each area is defined with its own archiving rules and procedures:

• Data ingestion of the ‘Submission Information Package’ (SIP): in this initial area of application, all the information about the material for archiving created by a producer during a production process using their preferred software environment, is brought together. Particularly in the case of computer-supported applications, this usually involves proprietary project data in software-managed filing structures which are not generally suitable for archiving and need to be exported to more neutral data formats. In the case of 3D models, archive formats are usually agreed on in advance, allowing them to be created as end products by the producer and transferred directly into the AIP. If the SIP is from an external source, it is reviewed by producer and management and recommendations for export are made. Once the material has been successfully stored in the AIP, the SIP can be completely deleted.

• Archival storage of the ‘Archival Information Package’ (AIP):in this second area of application, electronic data are preserved for the future. These data are never deleted but are continually upgraded by management to meet current technological developments using Life Cycle Monitoring. In the case of 3D data, emphasis is placed on representing geometry and texture in the highest possible resolution, in the most software-neutral and standardised format, with the longest possible lifetime. This ensures that, once collected, data are available in their totality for derivatives and future applications (requirement of Sect. 12.2.3).

• Data access of the ‘Dissemination Information Package’ (DIP):the third area of application involves collecting the data that were created along with the AIP, either in order to facilitate simpler day-to-day access (requirement of Sect. 12.2.1) or to allow it to be used with specialist applications (requirement of Sect. 12.2.4). Since the DIP is a derivative of the AIP, it is not archived long-term. This means that applications that are no longer readable or are technologically outdated can be completely deleted or replaced by newly created derivatives, allowing us to react flexibly to technological developments and to cater for fashionable, short-lived applications (requirement of Sect. 12.2.4).

• Data management with a ‘Document Management System’ (DMS): The fourth and last area of application does not have its own data package but manages the data units in the AIP and DIP. The DMS provides every item of data with descriptive additional details (metadata) about content, production, quality, archiving and much more. It also manages the allocation of data to particular information packages, which is vital for the correct implementation of the archiving strategy.

To summarise, the archiving strategy envisages the following data process: the producer generates the 3D model in the SIP and then exports it to the AIP, where it is permanently archived by management. At the same time, the producer always also creates a DIP for general access. If requested by a user, a special DIP can be created for data output from the AIP. If it is more convenient, specialists can also access the AIP directly.

3.4 Pilot Project 2021–2027

The strategy outlined in Sect. 12.3.3 will be adopted in the Archaeology Department of the Canton Zurich from 2021 onwards. There were several questions about the implementation and reliability that could not be satisfactorily answered in advance. For that reason, there will be an initial 5-year test phase, during which the strategy will be tried out and improved in the light of experience. The following questions will be crucial as we move forward:

• What data formats and data combinations will prove most successful as AIP?

• What data resolution will prove most practical for everyday use?

• What increase in storage capacity will be needed?

• How functional and amenable to further development is our chosen infrastructure?

In order to put the new system into operation, the workflow was operationalized using our Department software infrastructure (Fig. 12.5). We are currently using the software IMDAS Pro 6.0.18 by the Joanneum Research Institute with a standardised data model by the ‘spatz/imdas’ coordination group of the cantons of Grisons, Thurgau, Zurich and the principality of Liechtenstein. Therefore, IMDAS is serving as our data management tool and our storage system. All the sites, excavations, observations and findings are recorded in this system. At the same time, IMDAS has an integrated DMS where all the media object that need to be archived can be registered, along with their metadata, and linked with the relevant features. In addition to the existing media categories—‘image’, ‘plan/drawing’ and ‘document’—a new category has been introduced for ‘3D model’. Every 3D model is input as a separate media object and given a unique identification number. This means, that it can be linked to its associated feature (e.g. excavation or observation), while also serving as a dossier for all the relevant individual data files associated with the model (requirement of Sect. 12.3.1, datasets). These individual files are recorded in the DMS as matching media objects on their own. Only their storage location can vary depending on their purpose.

Fig. 12.5

Storage of 3D data in the current archive system of the Archaeology Department of Canton Zurich during the 5-year test phase

To implement the new system, it was necessary to specify initial obligatory data formats. These had to be compatible with Agisoft Metashape, the software used by Canton Zurich for the production of photogrammetry (in the current software version 1.4.4). For the duration of the test phase, we will be working with two mutually independent Archival Information Packages (Fig. 12.5, AIP Test Group 1 and Test Group 2).

Test Group 1 aims at an end product which provides the highest possible resolution, is software-neutral, and can therefore go on to be used in the greatest possible number of different applications—the criteria specified for AIP in the internal archiving strategy. To meet these criteria, in our opinion, a high-resolution mesh or dense point cloud offers the best solution, since all other formats can be derived from it. The output is produced using two industry-standard files: Wavefront OBJ for geometry and a Wavefront MTL file for texture (Agisoft 2020, p. 161; Library of Congress 2019). The package is completed by an automatically generated Processing Report, which summarises the parameters and error deviations used in the modelling process and thus allows the model’s sources to be critiqued (Agisoft 2020, p. 54).

In case the format chosen for Test Group 1 proves to be the wrong one, all the photos will also be temporarily archived. This will be done in a separate AIP, with the idea that the modelling can be repeated if necessary, supported by a computer-readable script that will allow the automated repetition of the most important modelling steps (Fig. 12.5, AIP Test Group 2). Agisoft Metashape offers a suitable, importable script in XML format (Agisoft 2020, p. 27), containing the final positions of the spatially oriented photos—or ‘cameras’, as they are called in Agisoft Metashape (Agisoft 2020, p. 19)—which are needed to compute the dense point cloud. Like most scripts, however, it is highly software dependent and time sensitive and will probably only be able to be opened directly in Agisoft Metashape for a limited time period. Test Group 2 is therefore not a real, long-term AIP solution. It is merely serving as a back-up system during the test phase and can hopefully be deleted in 2027.

All other files are optional derivatives, newly created, as described in Sect. 12.3.3, based on the AIP of Test Group 1, mainly using Agisoft Metashape. This leaves all doors open to provide users with an easily accessible and effective 3D experience. From 2021 onwards it will be possible, for example, to embed 3D models in blog posts using Sketchfab, or display virtual- or augmented-reality images via applications on smart devices in the context of public outreach work—just as requested in the first half of the paper.

4 Summary and Outlook

In this paper, the status of an on-going project at the Archaeology Department of the Canton Zurich is presented. This includes, for the first time, a concept for the archiving of 3D data. Photogrammetry is already an important component of excavation documentation and public outreach work in Canton Zurich, which will also become increasingly important in archaeological research in the medium term. However, the technology cannot yet be regarded as fully established, since data storage and archiving cannot be guaranteed. This should change by the time the project ends in 2027.

At the beginning of the project, a user-based analysis of requirements was carried out within the Department. This early canvassing of different points of view provided important insights for the subsequent formulation of a general archiving strategy. It turned out that when it came to using 3D models, a number of very different user groups presented different requirements with regard to data quality and data presentation (everyday users vs. specialists, public outreach staff vs. excavation staff). In response to this realisation, the possibility of multiple storage of information according to intended use was incorporated into the archiving strategy concept. This involved, in particular, separating archiving requirements from accessibility requirements. We adopted the Open Archival Information System (OAIS) as a suitable, up-to-date reference model, and this required us to define a clear distribution of roles within the Department, with relevant tasks and responsibilities. Preparations have also been made to put the strategy into operation. A five-year test phase for archiving 3D data will officially begin in 2021.

As this is one of the first ‘preservation plans’ for 3D data, we are about to submit our archiving strategy to the Swiss Centre for the Coordination of Long-Term Archiving of Electronic Documents (KOST 2020). So that others can benefit from our work. We will also report regularly on our experience in the recently created Swiss working group for digital excavation documentation (cf. DIG 2020). We hope that this article will lead to an international discourse to further facilitate the use, storage and re-use of photogrammetry.