This post comes from a paper I wrote for my Digital Archaeology class at the University of South Florida.
Archaeologists are rapidly embracing digital technology as a means to record, explore, and recreate cultural heritage with a high degree of accuracy and the added bonus of never having to risk an object’s conservation as the devices used never have to come into contact with it. Photogrammetry in particular is an excellent and affordable way to document objects, buildings, or whole landscapes using just a camera and some software. In this post, I will discuss the methods of photogrammetry as it is used to document objects and landscapes both from the ground and from the air. I will then use two case studies to show how useful it can be to archaeologists working with both ancient ruins and historic buildings. Finally I will suggest some challenges to the future of photogrammetry and the field of virtual archaeology in general.
Photogrammetry is a very new tool. It was mentioned within the Seville Principles, presented in 2010, as one of the new methods available to archaeologists to increase the scientific accuracy of their all-important task of documenting the world’s cultural history (López-Menchero Bendicho 2013). By the middle of this decade photogrammetry was being used for multiple scholarly projects (Olivito & Taccola 2014; Pierdicca 2015). Its benefits to archaeology, as well as its possible drawbacks, have been widely discussed in books and papers within the last three years (Jeffrey 2015; Olson; Olson & Placchetti 2015). As discussed in the case of the Vank Cathedral in Section 4, it has even been considered as a viable use of volunteers recruited through crowdsourcing platforms (Spanò, Hashemi, & Nourollahichatabi 2016). Most recently, photogrammetry has been incorporated into university classes designed to prepare students for using digital archaeology methods, such as one offered at the University of South Florida (Tanasi 2017).
Photogrammetry involves taking a series of photos which are then combined using software into a highly-detailed and textured 3D model. Objects, buildings, or even whole archaeological sites can be visualized quickly and cheaply by this method. Rather than needing expensive and complex equipment, photogrammetry requires only a digital camera and a software package to combine the photos, of which there are easy-to-use options available for free. The camera used to acquire the images should be at least five megapixels and the focal length of the lens should be no more than 50 millimeters. All points of the object being modeled must appear in at least three photos in order to align them correctly. Light should be uniform across the set or landscape and shadows minimized as much as possible. The software used to combine the photos struggles with very dark, reflective, or transparent surfaces, uniform patterns and solid colors, so objects with these qualities may not be good candidates for photogrammetry (Tanasi 2017, Nov. 25th). With the images acquired, tie points are extracted to align them and a point cloud is generated. Texture, taken directly from the photos, can be applied along with quality parameters such as the number of triangles in the model once the point cloud is converted into a solid mesh (Olivito & Taccola 2014).
For photographing small objects the camera should be shifted about 30 degrees around the object for each photo so that there is plenty of overlap. The background must be a different color from the object so that they do not blend into each other, as the background will have to be masked out of all the photos. It is important that the object does not move at all; it should be stable before shooting begins and everyone involved must be careful not to bump it or the table it sits on. Photogrammetric models do not provide an absolute scale; one must be added before work begins. Color and lighting should be uniform across the set and there is also the option to include color chips so that the photos can all be standardized later. Because the object must rest on something while the pictures are taken, it must also be flipped over and photographed again to get the bottom part; the two separate models are aligned later after both have been processed (Tanasi 2017, Nov. 25th). Since the object never comes into contact with the camera used to record it, as it would have using traditional molding techniques, there is no danger to it using photogrammetry (Balletti, Ballarin, & Guerra 2017).
The most straightforward method of capturing images for photogrammetry of a building or monument is to move around it or through it and take pictures of it from all angles, again making sure that there is 60-80% overlap in the images (Olivito & Taccola 2014). As shown by the case study of the Vank Cathedral in Section 4, this could theoretically be accomplished by volunteers with a little training working in their local areas with their own
cultural heritage (Spanò et al. 2016). Large buildings or landscapes may require alignment markers or georeferencing in order to orient the images correctly. The photographer should be careful of capturing his or her own shadow in the photos and try to work on overcast days or quickly enough so that the sun doesn’t move too much during the shoot (Tanasi 2017 Nov. 25th). Terrestrial photogrammetry can be used to rapidly document an archaeological site during excavation with a high level of detail and very little planning (Pierdicca et al. 2015).
With the advent of affordable remotely-piloted drones, aerial photogrammetry is becoming more available as a tool for the documentation of buildings or sites. Drones with cameras attached to them can automatically photograph a wide area with little or no intervention once they are set up. The use of drones for photogrammetry requires first planning out the flight, which should be done by an expert operator. Flights can follow gridded or circular patterns based on georeferenced images of the area to be photographed. Other parameters that should be set include speed, altitude, distance between waypoints, and the angle of sight and shooting mode of the camera. Camera choice is important, as attaching a camera that is too heavy will drain the drone’s batteries and limit flight time. Weeds or anything else that may affect visibility of the site from the air should be cleared ahead of time. An expert drone operator may be required for takeoff, landing, and any issues that come up involving wind, obstacles, or loss of signal, but otherwise the drone will automatically follow the pre-loaded settings (Olivito & Taccola 2014) Current trends in aerial photogrammetry include drones with high-precision GPS and automatic obstacle sensing that give them even further automation (Tanasi 2017, Nov. 27th).
Case Study: The Vank Cathedral
The Vank Cathedral in Iran was built between 1655 and 1664 by Armenian Christians who had immigrated to the area and others who had joined them having fled the Ottoman War. The church is built in a classic Persian mosque structure with double-shelled domes over a main hall. The interior of the church is decorated entirely with frescoes painted in a new style reflecting the combination of Armenian and Iranian cultures. Because of the importance of these frescoes, the focus of the project was to create a photorealistic reconstruction of them using photogrammetry (Spanò et al. 2016).
Another goal was to experiment with how such an endeavor might work if it was taken on using crowdsourcing methods and volunteers with limited training and access to equipment. In order to simulate such a situation, the team avoided the use of control points and any devices to raise the camera, instead taking precise measurements for scaling and shooting photos at three different vertical inclinations with a 30% overlap. Any photogrammetry project is highly dependent on the shooting strategy of the photographer and spatial position of the camera, but both concepts seem easy to grasp and perfectly teachable to non-experts (Spanò et al. 2016).
In order to create the model if the interior decoration of the Vank Cathedral, around 200 photos were taken using a 24 megapixel Nikon camera with an 18mm lens. The frescoes on the ceiling were captured by laying the camera directly on the floor pointed upwards. Chandeliers and clear protective coverings of the frescoes at visitor height caused problems; after some experimenting, the team found that the best way to overcome them was to split the project into two separate blocks of the main areas of the church. These two parts were then reunited by the strip of photos taken of the ceiling. The accuracy of the resulting point cloud was comparable to what would be expected using control points, with only slightly higher residuals (Spanò et al. 2016).
After combining everything in Agisoft Photoscan, converting the point cloud into a continuous surface, and applying the detailed images of the frescoes as textures using Technodigit 3D Reshaper, the finished model could be projected onto AutoCAD architectural drawings of the cathedral or “unwrapped” to show the painted scenes as a single, uninterrupted image. Taken together, the frescoes are arranged in five registers which read from right to left depicting lives of prophets, miracles of Christ, and the story of Saint Gregory, the founder of Armenian Christianity, among other scenes from the Old and New Testaments. By interpreting the paintings as a single image scholars are able to more easily focus on composition and iconography, as well as make comparisons between them and other artworks with similar themes (Spanò et al. 2016).
Case Study: Chan Chan
The site of Chan Chan in northern Peru is the largest pre-Columbian town made of mud bricks. It covers a very large area of fourteen square kilometers and was the capital of the Chimu culture. Nine palace complexes with walls and public ceremonial courtyards as well as more private inner spaces dot the area. The palace walls are decorated with bas-relief scenes of fishing and marine life, subjects that must have been important to people living just a few hundred meters from the Pacific Ocean (Pierdicca et al. 2015)
Chan Chan’s location is extremely dry and relatively cool, a situation that has provided for fairly good conservation of the earthen architecture. However, looting, salt air, and shifting weather patterns have done a significant amount of damage. As a result, many of the friezes are lost. The ones that remain have been documented only in photographs and are now protected (and hidden) behind new mud-brick walls. Combined with limited tourism possibilities and the sprawling nature of the site, it has been impossible to even see the important artworks, much less exhibit them. Technology is bringing new opportunities to study and display the cultural heritage of Chan Chan (Pierdicca et al. 2015).
In order to create a new augmented reality experience for visitors, the entry gate of one of the palaces, Palacio Rivero, was uncovered for a few hours of photography before being recovered for its protection.
Because of the necessary speed of the work, short planning timeframe, and lack of available equipment, the team decided to use terrestrial photogrammetry to create their 3D model. They accomplished this using a 24.3 megapixel Sony camera, with which they collected 440 images at a 40mm focal length with 30% overlap on each photo. The photos were taken from five points of view and then each set was joined into a separate panorama using stitching software in order to speed up the modeling and texturing processes (Pierdicca et al. 2015).
With the finished panoramas combined in Agisoft Photoscan, the resulting 3D model had almost 200,000 faces and 100,000 vertices. Such high resolution is great for archaeologists working to document or restore the structure, but the 134 megabyte file was much too large for consumer applications. By simplifying the model to just over 20,000 faces and 10,000 vertices, the file size was lowered to a little less than two megabytes, which is much easier for smartphones and tablets to handle.
The simplified model was then built into an iOS application that allows visitors to the site to display the model on their screens, projected onto the landscape in front of them. Thanks to the built-in GPS receiver of the camera that took the original set of photos, the model has terrestrial coordinates written directly into it that mobile devices can use to place the model correctly in its context. The gyroscope of the device can also orient the model in real time on the screen as the user moves around (Pierdicca et al. 2015).
Photogrammetry is quickly becoming an integral component of broader digital recording systems used by archaeologists. Although the value of using this method is well-established as a fast and accurate means to record data in the field, its scholarly value is still under some debate and will depend on how well it is integrated into existing systems of artifact analysis and recording (Olson). One thing that archaeologists must do in order to realize the full potential of photogrammetric models is to use them in innovative, collaborative ways. Otherwise they are just fancier, more accurate versions of the tools that are already in use (Olson & Placchetti 2015). More broadly, visualizations can only engage wide audiences and facilitate the exploration of an individual’s or community’s exploration of their own pasts if such tools make them feel connected to that past. Without that connection they will remain remote, sanitized tools for professionals and miss out on opportunities with the public (Jeffrey 2015).
With the ability to add more raw data to the record at a faster rate than ever and contributions coming from consumers and volunteers as well as professionals, managing all of that data is quickly becoming a problem. All of the original photos and scans, processed data, and finished models have to be kept, and multiple backups of everything have to go somewhere. Cloud storage services are readily available (and widely in use) but keeping projects online and repositories available requires funding. Connecting the public to the models means they must be decimated for viewing on less-powerful laptops and mobile devices, without losing so much resolution as to be useless. The perfect 3D viewer for researchers, one with plenty of tools and considered “academic” enough for scholarly work, doesn’t exist yet. These are all issues which the field of digital archaeology is going to need to tackle over the next several years if it is going to be able to truly use the new technology available to it (Tanasi 2017, Nov. 27th).
Photogrammetry is an incredibly powerful tool for recording and disseminating cultural heritage. Using this method of 3D modeling, archaeologists are now capable of recording their sites and the artifacts they excavate as they are working with a high level of accuracy and detail. They can then use the models to recreate an object, a building, or an entire excavation, and these models can be shared with fellow professionals and with the public. Photogrammetry is currently being used in projects all over the world, such as the Vank Cathedral in Iran and Chan Chan in Peru, and augmented reality provides the possibility of anyone with a smartphone or tablet to view a 3D model of a building on the landscape in front of them. There are some hurdles yet to cross, in particular ways to manage and best utilize the models and all of the raw data that goes into their production, but photogrammetry is undoubtedly going to continue gaining wider acceptance in a variety of applications.
Balletti C., Ballarin M., Guerra F. 2017, 3D printing: State of the art and future perspectives, Journal of Cultural Heritage 26,172–182.
Bruno F., Bruno S., De Sensi G., Luchi M. L. , Mancuso S. 2010, Muzzupappa M., From 3D reconstruction to virtual reality: A complete methodology for digital archaeological exhibition, Journal of Cultural Heritage 11, 42–49.
Jeffrey S. 2015, Challenging Heritage Visualisation: Beauty, Aura and Democratisation, Open Archaeology 1.1. https:www.degruyter.com/view/j/opar.2014.1.issue-1/opar-2015-0008/opar-2015-0008.xml
López-Menchero Bendicho V. M., 2013, International Guidelines for Virtual Archaeology: The Seville Principles, in Corsi C., Slapšak B., Vermeulen F. (eds), Good Practice in Archaeological Diagnostics, Springer, 269-283
Olivito R., Taccola E. 2014. 3D Modelling in the agora of Segesta: techniques and data interpretation, Archeologia e Calcolatori 25, 2014, 175-188.
Olson B. R., The Things We Can Do With Pictures: Image-based Modeling and Archaeology, in E. Walcek Averett, J. Michael Gordon, D.B. Counts (eds), Mobilizing the past for a digital future, The Digital Press @ University of Dakota, 237-250.
Olson B. R., Placchetti R. A. 2015. A Discussion of the Analytical Benefits of Image Based Modeling in Archaeology, in B. R. Olson, W. R. Caraher (eds), Visions of Substance. 3D Imagining in Mediterranean Archaeology, The Digital Press at The University of North Dakota, 17-26.
Pierdicca et al. R., Frontoni R. et al. 2015. Making Visible the Invisible. Augmented Reality Visualization for 3D Reconstructions of Archaeological Sites, in L. T. De Paolis, A. Mongelli (eds.), Augmented and Virtual Reality, Proceedings of the Second International Conference, AVR 2015, Lecce, Italy, August 31 – September 3, 2015, 25-37.
Spanò A., Hashemi N., Nourollahichatabi S., 2016. Image-Based Models Using Crowdsourcing Strategy, DigitCult – Scientific Journal on Digital Cultures 1.3, 65-79.
Tanasi D., 2017. Class Lectures, University of South Florida, Tampa, Florida.