By Daria Dobrochna Dabal
Photogrammetry is a real-world surveying technique that produces high-quality digital 3D models. An area is photographed in a specific way and the photogrammetry software analyses the photos to generate an accurate 3D model. The model is made by triangulating three-dimensional coordinates of every detail across multiple photographs.
Such 3D models are used by archaeologists to record artefacts or places of interest. During archaeological excavations photogrammetry can be used to record the site before, during and after excavation, showing each layer of context.
Because photogrammetry is a great surveying technique it’s slowly becoming an integral part of the archaeological toolkit. Whilst excavating an archaeological site we basically destroy a part of it. Documenting excavations layer by layer with all artefacts and archaeological features included offers a step-by-step record of the excavation works to portray research progress and the results.
As the popularity of photogrammetry increases, more 3D models of sites and monuments are made. But instead of using this technique to create better records of archaeological or historical features, a wide number of 3D models neither document nor record anything useful. Others show you archaeological sites or monuments in a way that doesn’t make any sense. All too often a photogrammetry model denotes the end of the process and can leave some wondering what to do with them. Most photogrammetry models end their lives as a small SketchFab window on the company’s website, where they gather dust and offer no real use. The main issues are lack of important details, context (surrounding/background) and errors of the models, missing bits and holes. Another import issue is finding a useful way to view and analyse the model.
It comes as no surprise that photogrammetry is currently not being used to its full potential in archaeology. Expectations from person to person vary, as do the skill levels. Photogrammetry software is being used as the final stage, where it should really be just the beginning.
Where most see the photogrammetry model as the end result, we see it as the beginning. We create the missing features, add animated objects and characters and build upon the information the model depicts. These models go on to become videos, interactive exhibits and even Virtual Reality locations, where you can walk around the site at one-to-one scale. We can even magnify the models in virtual reality. Imagine walking around an artefact that’s one hundred times larger, you really can see the details then. Imagine photogrammetry of an abandoned smithy, with its missing wheel digitally replaced and a water simulation flowing through it. Imagine looking inside the inner workings of that same, decrepit smithy and seeing it in action. That’s what we do.
Creating photogrammetry can be learnt quickly. We started by using the 30-day free trial of Agisoft Metashape, so needed to learn it within that month, to work out whether we’d be purchasing it or not. We watched YouTube tutorials and read the Agisoft Metashape User Manual. Our very first photogrammetry model was of the blast furnaces in Neath, UK. We only took around 20 photos with a mobile phone, which generated a 3D model that had giant holes and a distortion to its shape. Nevertheless, we were so impressed and excited with what the software can do that we decided to try to make better models.
This time we focused on improving the data collected from the site - better quality photos following a strict spatial capture pattern. We chose a much smaller site to practice on. Using a full-frame DSLR with a 50mm prime lens, as recommended in the manual, we photographed the Neolithic burial chamber at Parc le Breos on the Gower peninsular, UK. The model of this megalithic tomb looked brilliant. So we purchased the software and a drone and went to do more 3D models, to build up our skills in collecting data as well as using the software.
The experience of making 3D models taught us that capturing a good dataset is essential and, for what archaeologists need at least, the software does most of it for you. To create good looking 3D models that can be zoomed in and looked around you don’t even need to use all of the features that Agisoft offers.
To achieve a good dataset using a hand-held camera, as well as flying a drone, the right weather conditions are a big part.
Whilst working indoors you don’t have to worry about the weather but reflective surfaces like mirrors or shining metal objects, even glaze on tiles might cause a problem. Mirrors can be covered prior to the photo session and recreated in digital content creation software, like 3DS Max or Blender. Other interior elements and ornaments could be coated with pastes or powders to make them matte and scan well.
Use your camera with manual settings, keeping them the same throughout the dataset capture. Even slight changes in exposure may distort or cripple the generation of your 3D model.
Take photos methodically. The software works best if you present you photos in the order and pattern its expecting. Photograph every bit of the site, monument or artefact and ensure the images overlap each other. Bringing perfect photos from the site saves you a lot of time later when cleaning the model. Circle around the subject taking photos from different perspectives including crawling on the floor and taking aerial photos. Take photos with wide views and closeup. Also, if there are small or complicated details (graffiti, ornaments, sculptures etc.) treat them like a separate project by taking photos of them from a distance and close up.
Take more photos than you believe are required.
The thing to remember is, a 3D model is generated using a triangulation technique – seeing the same feature from many different angles and calculating its 3D position.
To build 3D models we use Agisoft Metashape, which is super easy to work with as it does most of the job for you. Alternative software is also available. You need to load the images and then the program will align them, working out where in 3D space each photo was taken. It then creates a sparse point cloud, which is a rough collection of dots that represents a 3D model, like a preview. The software will then build a dense point cloud, which takes quite a while and is dependent on how powerful your computer is. Our oldest computer took nearly a week to create the model. Our newest can do the same in a few hours. Eventually the dense point cloud is presented, which looks like your model and then you spend time deleting any points that are not needed or are incorrect. Then the software creates a polygon model and texture maps. The polygon model with texture maps are a standard format that can be exported from Agisoft and used elsewhere. The models are extremely detailed, with tens of millions of polygons.
Sadly, most people finish their work at this point and just deliver a project. So, in archaeology, most photogrammetry models end up as Sketchfab assets on a website, and that’s it. The majority of people find orbiting or hovering around the monument or artefacts as the best way (or only way) of presenting it. Some people render a flythrough using Metashape, which is usually lacking a floor and a sky, or any other elements which would help to understand the layout of the site. Such 3D models endlessly swirl around in emptiness and, unsurprisingly, people find the result underwhelming.
In general, archaeologists are stuck using Sketchfab, a website that hosts your 3D models for others to orbit. The models’ polycounts are so large that most laptops are crippled by them. Try to view someone’s model that boasted 32 million polygons on a smart phone – impossible. And Virtual Reality will be off-limits too. The original Oculus VR Headset can only manage up to one million polygons.
Of course, should you proceed further with your model, there are plenty of techniques to reduce the polycount to something more manageable by the masses. The techniques are known as Retopology, UV Unwrap and Texture Baking. You’ll need an experienced 3D artist to do this for you.
Retopology reduces the number of polygons required but keeps the original shape. UV Unwrap is a necessary step of mapping the model’s surface to a flat plane, allowing textures to be recorded. And Texture Baking is a process of transferring the original colour from your high poly model onto the low poly model. But that’s not all, all the extra tiny 3D surface details can also be projected onto your low poly model as a height map and normal map. These are maps that display tiny surface details on its surface, by raising or lowing areas. Computer games have been using this technique for years, giving Tomb Raider lots of detail with very few polygons, for instance.
If you have created photogrammetry and never got this far we have some good news - you can always do this stage later.
It is hard to predict how archaeological digital data and 3D models generated nowadays might be used by future generations. But it sure is fabulous to have a document which is a digital preservation or archive of the archaeological site which potentially could be available to everybody. It is exciting to think that one day everybody will be able to take a stroll through the excavation site or hold a virtual artefact in their hands.
Feel free to get in touch and discuss your past, present or future photogrammetry projects. We would love to offer you advice and discuss the possibility of working together.