This is the first in a series of posts exploring 3D modeling in Mediterranean and European archaeology. For more on this project click here. We hope these papers will start a discussion either in the comments of the blog or on Twitter using the #3DMedArch hashtag.
Brandon R. Olson, Boston University
Ryan A. Placchetti, University of Pennsylvania Museum of Archaeology and Anthropology
Recent developments in imaged-based 3D modeling have ushered in a new era, one in which the primacy of laser scanning as the chief means of three-dimensionally recording archaeologically relevant features and landscapes will be challenged. The release of a handful of image-based modeling software platforms, most notably Eos Systems’ PhotoModeler Scanner in 2009, Agisoft’s PhotoScan in 2010, and Autodesk’s 123D Catch in 2012, will transform how Mediterranean archaeologists plan, approach, record, and present their field projects and data for years to come. With the aid of Structure from Motion (SfM) and other comparable photogrammatic algorithms, it is now possible to create accurate and photorealistic 3D models of any target of interest using digital photographs (Figs. 1 and 2). Such a technological breakthrough, however, though capable of transforming archaeological field methods, could, if not utilized correctly, hinder the discipline as well. The purpose of this investigation is to examine the utility and limitations of image-based 3D recording from an archaeological perspective before discussing how 3D recording can become a legitimate analytical tool for the archaeologist, rather than just another means to generate visual aids.
The Utility of Photogrammetric Approaches to 3D Modeling in Archaeology
A number of recent studies have demonstrated that image-based 3D modeling, when the target is prepared and photographed correctly, provides aesthetically pleasing and spatially accurate 3D renderings of archaeological features ranging in size from individual artifacts to entire landscapes (Olson et al. 2013; Remondino 2013; de Reu 2013; Verhoeven 2011). In fact, field tests of PhotoScan Pro revealed that the software can generate models with up to two centimeter spatial accuracy across a 25 sq m space (Olson et al. 2013; de Reu 2013). The utility of the technology rests not only with its accurate outputs, but also in its affordability and ease of access. 123D Catch represents an open source option, but even proprietary modeling software packages are cost effective with PhotoScan Pro offering an educational license for $549 USD. Although certain programs are more user-friendly than others, the programs make efforts to present easily navigable user interfaces, especially 123D Catch and PhotoScan Pro. Irrespective of one’s technological background, one can achieve a basic level of functional literacy in an afternoon with training from an experienced practitioner. Archaeologists familiar with ESRI’s ArcGIS software will appreciate the comparable lack of expertise required to master the full range of functionality offered by 123D Catch and PhotoScan Pro. The aforementioned image-based 3D modeling programs offer something that very few archaeologically appropriate technologies do not, they are cheap, easy to use, portable, and yield quality outputs. As such, archaeological projects throughout the eastern Mediterranean such as the Tel Akko Total Archaeology Project (Olson et al. 2013; Killebrew and Olson 2013), the Central Lydia Archaeological Survey, the Athienou Archaeological Project, Qazion (Quartermaine et al. 2013), the Jezreel Valley Regional Project, the Pyla-Koutsopetria Archaeological Project, Troy, «atalhˆy¸k, Polis-Chrysochous among many others, have implemented the software to address specific needs.
Figure 1: Image of a secondary apse from a Late Roman basilica at Polis-Chrysochous, Cyprus depicting the three stages of creating a 3D model using an image-based modeling software: A) The automatic alignment of photographs; B) A 3D point cloud; C) The untextured 3D model; D) The final textured model. Thanks to Bill Caraher for permission to design and present this image.
Limitations and Challenges
The current suite of photogrammetric 3D modeling platforms offers an appealing technological addition to any archaeological project. These software packages, however, present a series of specific and broader methodological limitations. The corpus of specific limitations has been discussed elsewhere, but a few select drawbacks bear mentioning (Olson et al. 2013; de Reu 2013). A great deal of the frustration and false starts associated with incorporating photogrammetric 3D modeling into an archaeological project can be avoided with preparation. First of all, developing a fundamental understanding of how SfM technology reconstructs a spatial environment across a series of photographs, combined with familiarity of proven photographic collection strategies can help to inform the tailoring of a data collection plan for a specific modeling target. Secondly, 3D modeling may not be appropriate to all objects and areas due to technical limitations of the software. Scenes with strings or grass, monochromatic cylindrical objects that lack clearly defined edges when viewed in the round, as well as glass and other transparent surfaces are all problematic features that prevent optimal modeling. Finally, photogrammetric 3D modeling operates best in a controlled environment, but it is perfectly viable in field operations provided that steps have been taken to minimize problematic features and to standardize the scene across the photographic dataset to maximize the computer’s efficacy when trying to recreate the spatial environment.
The broader methodological limitations of the software concern its use by the archaeologist. Utilization of object-based 3D modeling must be approached with purpose and consideration to ensure that the data are utilized in such a way that augments data collection instead of distracting from it. Understanding the investment of effort required to produce and manage a viable 3D dataset requires a basic understanding of how the technology recognizes spatial relationships and the volume and type of data produced at each step of the modeling process. Most importantly, it is necessary to have a clear plan for the role that these models will play within the scheme of a specific archaeological project. In the absence of proper planning, 3D models run the risk of being relegated to curiosity status, functionless byproducts of an overeager adoption of technology or worse, particularly in the context of an ongoing excavation, they might prove to be a waste of valuable time and information if the final product should prove unsuitable for use.
Future Directions: 3D Modeling as an Analytical Tool for Archaeology
As projects continue to adopt this technology, it is important to reflect on how and in what ways image-based modeling can become an analytical tool, as opposed to a means of simple archaeological visualization. What is presented below are four avenues that we believe would benefit from quick, accurate, and photorealistic 3D models, and their 2D data derivatives. It is important to note that discussions with a number of colleagues, including Bill Caraher (University of North Dakota), Christopher Roosevelt (Boston University), Jody Gordon (Wentworth Institute of Technology), Ann Killebrew (Pennsylvania State University), and Curtis Runnels (Boston University), helped shape and contextualize what follows.
The documentation and dissemination of spatial data is one of the most common concerns of field archaeologists. Traditionally, cartographers seeking to map archaeological features set up a series of datums across an area to be mapped and take a number of measurements in order to produce a hand-drawn map of a feature or site. The process is time intensive and the accuracy of the map is dependent upon the tools used and the skill set of the illustrator. In many cases, upon completion, the map is scanned and opened in a graphics editing program or GIS for digitization. With respect to site and excavation unit level mapping, the benefits of a digital approach are evident, it is faster and more accurate than a manual approach. PhotoScan Pro and PhotoModeler Scanner offer a 2D georeferenced orthorectified photograph (henceforth referred to as orthophoto) output where a 3D model is converted to a spatially accurate 2D rendering of the modeled space. The images serve as an ideal basis upon which accurate maps can be digitally drafted.
Field testing has demonstrated that maps of scales ranging from an excavation unit to an entire site created with orthophotos from an image-based 3D modeling software are more accurate than maps created with manual drafting methods (Olson et al. 2013; Olson and Killebrew forthcoming; de Reu 2013). Olson and Killebrew note that maps of excavation units from Tel Akko were drafted using an orthophoto with sub-millimeter resolution and spatial accuracy averaging 2 cm, while the 3600 sq m site of Qazrin was digitally mapped using an orthophoto with 5 mm resolution and 7 cm accuracy. The two case studies prove that a digital cartographic approach predicated upon the use of orthophotos exported from an image-based 3D modeling software provide a time efficient approach to archaeological mapping with unprecedented spatial accuracy.
Field Recording and Volumetrics for Archaeological Excavation
Current documentation strategies of ongoing excavation most often takes the form of paper forms, narrative description, and spatial recording in a geographic information system, which are all 2D based methods that seek to record a 3D space. Attempts to develop a 3D documentation system for excavation have been undertaken with varying degrees of success (Gidding et al. in press; Katsianis et al. 2008; Sanders 2011; Smith and Levy 2012). Despite such attempts, an ideal system, one that is photorealistic and spatially accurate and designed to store excavated data in a 3D environment that enables calculations of volume, an examination of spatial relationships, and is easily updatable, does not exist. It is clear that a digital geographic environment that is spatially controlled and capable of displaying textured 3D data has yet to be developed. ESRI’s ArcGIS 10.1 software and ArcScene do not yet support such an environment. Despite these limitations, the spatial integrity of image-based 3D models would serve as an ideal basis for a 3D excavation recording system, given that most of the software packages permit area and volume calculation functions.
In the past, object analysis has been reliant on either direct physical access to objects or the generally inferior experience of pouring over 2D representations or written descriptions. High-fidelity 3D digital artifacts can help to bridge the gap in quality of data available to researchers by improving the experience of indirect object interaction. Even the physicality of remote object interaction can be partly addressed by incorporating 3D printing technology. The relative immediacy with which digital artifacts can be disseminated across physical boundaries and distances will prove beneficial to scholarly collaboration, while conservation considerations of physical storage space, safe transport, and object deterioration are a non-issue. In a matter of hours, an object discovered in the field can be sent as a 3D model anywhere in the world, expanding the pool of expertise beyond those present at the excavation site to provide a more informed initial analysis.
Figure 2: Cutaway showing the point cloud, untextured model, and textured model of an Acheulean handaxe. Thanks to Curtis Runnels for granting access to this artifact.
The act of creating a 3D model is a step towards digital preservation. Digital records, provided they are kept up to date with current file format standards, do not deteriorate over time. Rather, they are made resilient by their transferability, duplicability, and finite dimensions. Unlike the locations and objects being recorded, a 3D model is a static representation of its subject at a particular moment in time. Future events that alter the original, whether deleterious or beneficial, make no change in the record. In this sense, the photorealistic 3D model is the documentation tool available to archaeologists that most closely approximates the experience of the original. While the digital replica is not a truly equal substitute, it can help to mitigate the loss of information and context caused by the forces of nature and destructive human activities such as construction, vandalism, or even further excavation. In extreme cases, where the subject of a model is going to be destroyed and time is limited, as in the case of many rescue archaeology projects, 3D documentation is a fast and reliable way to capture the ephemeral final moments of an archaeological find. In more ideal conditions, the model can also serve as a visual milestone in the life of a subject’s archaeological development, providing a record prior to reconstruction or the implementation of protective measures, preserving the find in its most original state and stripped of recent artifice. The opposite also holds true, should the subject of a model fall into disrepair, a 3D record can also provide a basis for restorative work. A 3D model is no replacement for the authentic experience of an original, but it can potentially serve as an enduring record of an artifact, feature, or site in a field burdened, even under the best conditions, by the inevitable degradation of material over time.
As modern archaeologists, we are expected to be thorough and effective custodians of the information with which we have been entrusted. Image-based 3D modeling is new, exciting, and scientifically valuable, but should be approached earnestly with clear analytical goals in mind. In terms of realism, 3D models may offer vastly superior visual and spatial representations of objects and areas than traditional 2D methods, but they also represent an analytical resource, provided that diligent consideration of how and why a 3D dataset is created takes place before the first picture is snapped.
The potential to improve mapping methods, spatial recording, object analysis and access, and digital preservation has the potential to transform the discipline. Archaeology is a destructive act because all human works are temporary and attempts to faithfully record and duplicate objects and areas in a digital environment are the closest an archaeologist can come to recreating the moment of discovery. Ready access to 3D modeling provides a reasonably satisfying facsimile of a real world subject when called for, and making permanent what is by definition a temporary state of existence.
The ability to faithfully record, digitally duplicate, and rapidly disseminate photorealistic 3D representations of subjects of archaeological interest is only possible when approached with foresight and only valuable if researchers in the field and in institutions find ways to create effective collaborative spaces. In the absence of collaborative innovation, the archaeological field runs the risk of simply mimicking the results of traditional tools and methods, granted, faster and more accurately, but without realizing the full potential of a digital recording system.
Brandon R. Olson is an Adjunct Instructor at Tufts University and a Ph.D. Candidate in the department of archaeology at Boston University.
Ryan A. Placchetti is a Research Associate at the University of Pennsylvania Museum of Archaeology and Anthropology.
de Reu, J., G. Plets, G. Verhoeven, P. De Smedt, M. Bats, B. CherrettÈ, W. De Maeyer, J. Deconynck, D. Herremans, P. Laloo, M. Van Meirvenne, and W. De Clercq. 2013. “Towards a Three-Dimensional Cost-Effective Registration of the Archaeological Heritage,” Journal of Archaeological Science 40: 1108-1121.
Gidding, A., Y. Matsui, T. E. Levy, T. DeFanti, and F. Kuester. In press. “ArchaeoSTOR: A Data Curation System for Research on the Archaeological Frontier,” Future Generation Computer Systems.
Katsianis, M., S. Tsipidis, K. Kotsakis, and A. Kousoulakou. 2008. ”A 3D Digital Workflow for Archaeological Intra-Site Research Using GIS,” Journal of Archaeological Science 35: 655-667.
Killebrew, A. E. and B. R. Olson. In press. “The Tel Akko Total Archaeology Project: New Frontiers in the Excavation and 3D Documentation of the Past,” Proceedings of the 8th International Congress on the Archaeology of the Ancient Near East. Wiesbaden: Harrassowitz.
Olson, B. R., and A. E. Killebrwew. Forthcoming. “A Photogrammatic Approach to Digital Drafting in Archaeology,” Journal of Eastern Mediterranean Archaeology and Heritage Studies 2.
Olson, B. R., R. A. Placchetti, J. Quartermaine, and A. E. Killebrew. 2013. “The Tel Akko Total Archaeology Project (Akko, Israel): Assessing the Suitability of Multi-Scale 3D Field Recording in Archaeology,” Journal of Field Archaeology 38: 244-262.
Quartermaine, J, B. R. Olson, and M. Howland. 2013. “Using Photogrammetry and Geographic Information Systems (GIS) to Draft Accurate Plans of Qazion,” Journal of Eastern Mediterranean Archaeology and Heritage Studies 1: 169-174.
Remondino, F. 2013. “Worth a Thousand Words- Photogrammetry for Archaeological 3D Surveying,” in R. S. Opitz and D. C. Cowley, Interpreting Archaeological Topography: 3D Data, Visualisation, and Observation. Occasional Publications of the Aerial Archaeology Research Group 5. Oxford: Oxbow, 115-122.
Sanders, D. H. 2011. ”Enabling Archaeological Hypothesis Testing in Real Time Using the REVEAL Documentation and Display System,” Virtual Archaeology Review 2: 89-94
Smith, N. G., and T. E. Levy. 2012. “Real-time 3D Archaeological Field Recording: ArchField, an Open-Source GIS System Pioneered in Southern Jordan,” Antiquity 86: http://antiquity.ac.uk/projgall/smith331
Verhoeven, G. J. J. 2011. “Taking Computer Vision Aloft-Archaeological Three-Dimensional Reconstructions from Aerial Photographs with PhotoScan,” Archaeological Prospection 18: 67-073.