In 2005, we received a small grant from the ![[WWW]](http://wikispot.org/wiki/eggheadbeta/img/sycamore-www.png){kind=small}
Agnes Cole Dark Fund to explore the possibility of using photogrammetric techniques with historical photographs. The process involved scanning historical photos and using them to reconstruct 3D models of historic sites. In some cases there was enough information in the historical photo to work with; in other cases, it was necessary to combine historic and present-day photographs of the same place.
For our prototype work we used PhotoModeler Pro software. Schematically, the process is relatively straightforward. Starting with a digitized photograph, a three-dimensional space is created, as suggested in following image.
The structures that still exist in the present-day world can be measured to provide a metric for this space. This means we can figure out the size and position of structures which no longer exist, like the building coloured green in the image below.
Having such a space means that it is possible to superpose historical and present-day images, as in the following composite. (Superposition simply means putting more than one thing at a time in the same place or the same space. We use the term to highlight the fact that it is often possible to learn something new when you bring representations together, or when you use representations together with place.)
It is also possible, with enough historical and present-day photographs, to build more sophisticated 3D models using the same techniques. At present, this kind of work is very painstaking and time-consuming because it has to be done by hand.
Our hope is that these kinds of techniques can eventually be automated and scaled-up. Imagine being able to automatically match historical photographs with images of present-day places to determine where they were taken. Such a system would require access to a large database of present-day photos whose exact location is known (in other words, to geocoded photos). We are currently in the process of experimenting with just such a database. The system would also depend on software that can automatically reconstruct 3D properties from two-dimensional scenes. Techniques for doing this are currently under development at a number of computer science laboratories.
One thing that we did learn during the project is that there are steep social and institutional costs associated with trying to use photographs that have not already been digitized. Future projects will focus on photographic databases that are already in digital form.
