The Automated Detection of the Roman Army in the Netherlands

Bison code: L.28296

Almost 2000 years ago, the border of the Roman Empire ran along the river Rhine, straight through the Netherlands: the south was Roman and the north was barbarian territory. However, this wasn’t just an administrative ‘line in the sand’, there was Roman military activity everywhere: we know of some 20 fortresses, but many other military elements are hard to detect because of their size (watchtowers) or their temporariness (expedition and marching camps). Furthermore, the infrastructure of the Roman Army is completed by roads, ramparts and canals. Until now, only a handful of the above mentioned military installations have been discovered.

However, to understand the exact workings of the Roman border system (the Limes), we need to identify as much military constructions as we can. There are two challenges with this: a) these constructions are, of course, hidden beneath the ground, and b) they can be situated in large parts of the (central) Netherlands.

In recent years, the summers have been extremely dry. This has given us a unique opportunity because, in some cases archaeological remains, such as military installations, can suddenly become visible as distinct patterns in the drying out of grasses. By studying aerial imagery of these dry summers it is possible to locate this kind of archaeology quite clearly. The challenge then, is to manually go through an enormous amount of aerial images, since these military installations can be situated in such a wide area.

The first task for this project is to compile a set of characteristic features of Roman constructions in freely available aerial imagery. For the compilation of this dataset, the results of a German project in a comparable landscape can be used. The next step is to investigate and to test if it is possible to develop a computer vision system that uses image processing techniques and machine learning to automatically identify these Roman military features. The last step is to find a way to search a large amount of freely available aerial imagery on the internet for these type of features.

Cluster: Remote Sensing and Soil Measurements 1