Maps were a key technology in modern navigation yet, until recently, the quantitative study of historical map’s content on a large scale has been limited by constraints on access to materials and by computational and technological limitations. Consequently, existing historical studies dealing with cartography have relied on representative examples and curated comparisons, without engaging in formal large-scale investigations. The recent flourishing of new digital technologies and materials encourages different approaches. In line with recent applications, this contribution presents a new digital method to automatically georeference and register changes in historical maritime cartography.
Currently, georeferencing is almost invariably done “by hand”, with the user imputing specific control points on digital raster images. The control points are associated with the equivalent points of known coordinates on the globe. Existing algorithms can then, with increasing accuracy as the points increase in number, create a georeferenced raster that is readable by a GIS software. This process, however, can be very time consuming.
The approach proposed in this contribution, by using a multi-step method that combines deep learning techniques and image analysis, automates the procedure with promising results, and, leveraging the statistical flexibility of deep neural networks, can work on maps characterised by heterogeneous styles. The procedure offers a rapid path to geographically position families of map scans for further analysis.
An example of the type of historical analysis supported by this procedure is carried by using a segmentation network and error metric to characterise the accuracy of each map. This is determined by proxy, using the distance of 21st century coastlines from their historical equivalent as reported on the digital map scans. In this way, a dataset can be constructed that compares maps across regions and producers. They are comparable through a constant framework enabled by the fact that, since the 17th century, longitude and latitude information and standards were firmly established in mapmaking. One of the benefits of such an approach is the construction of a repeatable assessment of maps as seen in their technological dimension, defined here as the quality of their spatial information.
A test dataset and analysis obtained with the technique is presented for the years 1650 to 1750 AD.
The approach offers insights into the methodological challenges around digitisation, as well as the rewards and malleability digitisation affords. It is also an example of how established and diffused sources can be connected and studied in new ways.
Giovanni Pala is a final year DPhil candidate in Economic and Social History at Oxford University. His dissertation explores the evolution of cartographic accuracy ca. 1650-1850 as a measure of the quality of geographic knowledge and performance of maps. His interests are in the History of Technology, the Economics of Knowledge and Culture, and the use of Digital Scholarship methods to process non-textual sources.