The 3 requirements for 3D reconstruction
For digital historical reconstruction three requirements need to be fulfilled:
- The source must be copied with all relevant details
- The copy must be reproduced with a high level of realism
- The copy should live on in a similar way as the original
A lot of focus traditionally in on the second requirements: almost every week new hardware and demos are presented. They all try to increase the level of realism in which a virtual environment can be experienced. Palmer Luckey, the founder of Oculus “within ten years we have a good shot of creating virtual reality technology that allows you to make an experience that is nearly indistinguishable from reality.”
But then the other two. To start with the first: it took more 1 year of my spare time to recreate the model for Hummelo from old photographs and even now I know there are many area’s where things are not quire right or where I simply do not have the material to create a correct model. So in the long run, we should try to find new ways of making a full digital copy of (past) reality.
We tackled this in 2D around 2000 years ago with the invention of photography it was possible to make a copy of reality and look at it at a later moment. We have progressed a lot since then, but we still do not have methods for 3D. We now have 360° videos, 3D video and TVs, but neither is an actual 3D model. The viewing position is fixed and in the latter depth information is not present. So although makeview has a Oculus Rift version of some of their videos, you will not have a stereoscopic view.
But recently we see progress in this area: Google no longer thinks hand-modelling houses for Google Earth is the way to go. Instead the are generating the 3D models from airplane fly-overs. Just like the now well-known Streetview cars taking 360° photos every 100 meters. But in this case the base material is processed such that 3D information is retreived and the photo information is projected on it. It is still a bit like the photo matching in Sketchup, but then automatic. I expect this technique will mature very fast in the forthcoming time and that making a 3D model of an area will possibly be as simple a driving a Streetview v2.0 car around.
The last requirement is the hardest of them all, because that means that we need to copy the invisible properties of the environment: the behaviour of the people and nature. We may even need to have a kind of evolution. Getting this all modeled is very difficult if possible at all. The techniques needed for this need to come from the area of Artificial Intelligence. In the first half of the 90-s I have studied AI at the university. My conclusion back then was that this was still in a very primive stage of development and I haven’t changed my conclusion since then. In this case however, we do not need to generate human bahaviout, but to copy it. This is a a slightly different angle to the same problem and perhaps this will give directions for development.
I would love to take up the challenge in fulfilling the 3 requirements above. I think there is a lot going one which creates unprecedented possibilities. But for now, this all is just an hobby.