Skip to main content

Forensic Architecture: using 3D modelling for the people

A group constructing an accurate model of violent events using 3D imaging based on multiple kinds of input can help us discover the truth of situations the powerful want to keep hidden, write ROX MIDDLETON, LIAM SHAW and JOEL HELLEWELL

ALL models are wrong, but some are useful. This aphorism is generally attributed to a statistician in the 1970s, but its message is much older, and its truth is much more widely applicable than just to statistical models.

All representations of the world are “models,” that is, depictions of reality, or some aspect of it. They are “wrong” because they purposefully simplify reality — to make it comprehensible, or to extract a perspective through one particular lens.

When violence is inflicted on people by the powerful, they work very hard to control the story of what has happened. They offer an account of the world, a model, to describe what has happened on their terms.

This simplified model generally involves hiding or “losing” evidence, discrediting or silencing witnesses, and making false claims. To fight the loss of a good description of the truth of what happened, better evidence, better information and better models are needed.

One of the most powerful forms of model is a visual image, or perhaps better, a 3D visual representation, from which images can be extracted from any perspective. The explosion of camera availability over the past several decades means that the world we live in has become supersaturated with visual data.

Millions of cameras record daily such immense volumes of data that no-one will ever, or could ever, process it, even if they had access to them, which for the majority of cameras, owned privately or by states, they don’t. We are living at a moment in history in the early stages of creating information from such large volumes of data. One way is to use these images to produce 3D models.

The field of photogrammetry develops techniques to create 3D models from light-based imaging and is booming, along with the more frequent use of 3D models. These new techniques are revolutionising fields from archaeology to forestry. The technology is also rapidly making its way into consumer electronics.

One of the most successful 3D mapping techniques uses a laser pulse to scan the surroundings. The time between the laser pulse being sent from the device and it being reflected from an object and arriving back at the source is used to tell how far away the object is.

Measuring almost unimaginably short times (for something five metres away, the time for the pulse of light to return is 30 billionths of a second) now poses no technological challenge. The technology is called LiDAR (analogously with radar) and has already been introduced across several mobile phone models.

The other method of recording 3D images uses a more intuitive and apparently simpler approach. Just like our eyes, multiple cameras a fixed distance apart can record the small differences in perspective that are required to understand the 3D shape of an object. Multiple cameras are also common in mobile phones.

But as we’ve noted, the lack of images isn’t really the problem in making 3D models. We are drowning in a surplus of images, from the surveillance industry, to the self-recording and counter-surveillance that individuals can now create, especially when something important starts to happen. Footage from stills and videos is uploaded by people all over the world, including in war zones and under extreme duress.

What’s still hard is the stitching together of images taken from different perspectives, times and devices. When the relationship between two cameras isn’t known, stitching images requires anchor points — multiple recognisable and distinct details that are visible in all images, and from which the cameras can be related to each other, as well as the objects in the scene.

The most familiar might be the ping-pong ball studs of a motion-capture suit worn by actors performing the actions that will be converted to animated characters in movies.

These anchor points are hard to identify, but they should be present in almost any scene. The Forensic Architecture research group based at Goldsmiths, University of London, uses buildings, urban environments and landmarks to do just this.

They take publicly available recordings like videos, photos and voice recordings, as well as witness accounts, maps and other data, to reconstruct events relating to “human rights violations including violence committed by states, police forces, militaries, and corporations.”

These are the sort of important events that people increasingly try to capture on camera. The researchers identify anchor points to stitch multiple accounts of violent events together through the use of 3D models of the locations where they happened.

The focus is not on general automated 3D models as a technical challenge, but on the accurate reconstruction of critical moments to extract important information. These models are then used in interviewing survivors of the events and experts in relevant technical and legal fields. Their work has been used in courts as well as in journalistic investigations.

The work they do is visually impressive as well as clear, informative and persuasive. The way in which the world is recorded and shared is so compellingly interwoven, that their work belongs as naturally in a court as in an art gallery (or nominated for the Turner Prize) or through downloading their publicly available models and datasets.

Forensic Architecture’s methodology is fantastically powerful and should become widespread as the future of image aggregation. They have tackled a wide range of subjects, from the killing of refugee Muhammad al-Arab at the Greek border, illegal forest burning by a palm-oil conglomerate Korindo, the origin of the Beirut Port explosion and the death of Adama Traore in police custody in France.

Their multidisciplinary approach shows how we can use the tools we already have, and the power of citizen journalism, to uncover the secrets and obfuscations given by the powerful. The range and depth of their investigations are breathtaking and show that in a world saturated with data what the value of a useful model really is. Visit Forensic-architecture.org to view their work.

 

OWNED BY OUR READERS

We're a reader-owned co-operative, which means you can become part of the paper too by buying shares in the People’s Press Printing Society.

 

 

Become a supporter

Fighting fund

You've Raised:£ 12,411
We need:£ 5,589
5 Days remaining
Donate today