Modelling the evolution of frontiers

Understanding and presenting the evolution of frontiers, where digital maps can help tremendously, allows us to draw conclusion on the cultural background of the studied area. But digitizing historical maps is a very cumbersome undertaking: After scanning, georeferencing and rectifying a map, the objects to extract have to be traced accurately. In order to reduce the time spent, we want to shorten at least the last step of this process. Therefore we have decided to develop a QGIS plugin for semi-automatically extracting frontiers from a georeferenced map. With the assistance of the user, our algorithm shall find frontiers and digitize them.

Using our plugin, we will digitize a set of input maps, showing preferably the last couple of centuries of the Republic of Venice, and with these create an animation that shows the evolution of the frontier over time. This will serve as an example as what can be done with our plugin.

Our milestones divide the project in three phases:

  • Preparation
  • Implementation
  • Creation

GANTT M3

Preparation

In this first phase we want to get ready for actually working on our project. We want to start with collecting an initial set of sample maps. This step is very important because it allows us to get an impression of how the maps, with which we are going to work with, really look like. Our idea is to go to a library and search for historical atlases. As soon as we have found enough sample maps we want to scan them. Then, being in the digital form, we’ll have to manually georeference them. This task won’t be easy. Historical maps aren’t know for being very accurate and we’ll have to deal with distortions and inaccuracies, where the distortions can to some extent be resolved using a QGIS-plugin for rectification. We have calculated three weeks for this step because we will possibly have to wait for a book to arrive at our library.

At the same time we want to set implement a “Hello World”-QGIS plugin. By doing this we get to know how such a plugin has to be implemented and which interfaces with QGIS exist. Further we can use this first plugin for testing purposes. Already we have found several tutorials on the web that show how to implement a QGIS plugin.

The last point in the preparation phase is the setup of administrative things. For version control we are planning to use a GIT repository either on the server of the EPFL or on another free service. Further we want to maintain a log book for keeping track of what we are doing during our project. We thought that Framapad might be a suitable platform for doing this.

The end of the preparation phase is marked with our first milestone after 4 weeks. Summarized, what we want to have at this point is: scanned sample maps, a simple QGIS plugin structure and all administrative services set up. Having all this, we are ready for the next phase.

Implementation

The four weeks lasting implementation phase is the main part of our project. At the latest after the first milestone we want to start the research for methods and algorithms for detecting frontiers on a scanned map. We will have to test and compare several methods for being able to choose the most suitable.

Then we will integrate the chosen method in a QGIS plugin such that it can be operated in a user friendly way. What exactly the term ‘user friendly’ means in our application and how we will implement the user interaction will be strongly influenced by the choice of image processing algorithm.

The budgeted 4 weeks for this second phase are very optimistic and will probably not be met. But we’re of the mind that a good plugin would strongly facilitate the last part of our project, the creation of a representation of the evolution of frontiers. Further it could probably, as a future work, easily be expanded for detecting other objects on a map (e.g. houses on a Napoleonic cadastre). We therefore see the plugin as the most valuable output of our project and it is thus our main deliverable. Hence we will really focus on the successful termination of the implementation phase. As a summary, at the end of this implementation phase we will have a working plugin that semi-automatically finds borders on a scanned map and digitized it onto a manually georeferenced digital map.

Creation

In the last week, we will use the QGIS-plugin that we wrote to process a couple of maps, so that we can then create an animation or something similar. The source of these maps will be the same as in the preparation phase. We hope that we can find enough sources to show the evolution of the frontier of the Republic of Venice over many of the centuries of its existence. How the end result will look is still open and anything from a video, a Google maps overlay (like our predecessor did last year) or an animation in some other form is possible.

And last but not least, in this final phase, we will also finish the final presentation and final report for the project.