The aim of the Generalisation team is to automate the derivation of products (maps or datasets) from the large scale data that Ordnance Survey collects and maintains. Full automation is often only inspirational, the real aim is to find the most cost effective way to create and maintain products. It usually involves a balance of manual and automatic processes, and this balance continually changes. We gradually increase our expertise and improve our toolset for automating the processes, which in turn allows us to envisage the creation of more challenging products, requiring more complex automation.
The team pursues two types of activities, aiming either at providing short term benefits to Ordnance Survey, or looking a bit further to explore the opportunities that the advances in the domain or related domains will bring to Ordnance Survey.
Reducing the cost of map production
After a few years spent developing a rich library of software tools to perform generalisation tasks, the generalisation team is now in the position to create complex automatic processes that can derive maps automatically from our large scale data. While the graphic quality does not quite match what can be done by experienced cartographers, the quality is still high enough for contextual products (maps used as backdrop for the users to add their own data on top), and the production cost is much lower. For example the initial version of OS VectorMap District took less than 4 months to be created, from design to delivery of the country-wide series of maps. This would not have been possible without the automatic generalisation processes provided by us.
We are now taking part in a number of projects aiming at redesigning some of our cartographic production systems to make them more cost effective.
On demand mapping
While feeding into the design and development of new map production systems, we are also looking into the future, trying to understand what increasing automation might bring to Ordnance Survey and its customers. We believe that increased automation will allow us to move from a situation where we propose a small range of general purpose products, to a wider range of more targeted products. This would not be possible today, as the cost of production of these specialist products could not be recovered from the sales to a very limited customer base. But if we can build automatic map production systems that can take customer requirements into account, then it will become viable and will bring great benefits to our customers, by providing them with products tailored for their needs. We are still a long way from this though. The challenges are to gather the user requirements, express them in a machine understandable way, and then design a system that can use these requirements to select the appropriate data, and perform the appropriate transformation to bring them to the appropriate level of detail, ensure consistency between the layers, and deliver a cartographic representation that meets the requirements.
Requirements to support on-demand mapping
The first large project that we have undertaken in the context of the research into on-demand mapping is to propose the architecture for such a system. We need to understand what knowledge the system will require; how it can be represented; and how to establish the connection between knowledge that comes from the user domain (application domain), and knowledge required by the system to retrieve data; decide what transformations need to be applied; and find the tools to actually perform those transformations on this particular data. The aim of this project is therefore to identify all the components of such a system, and the standards that are required to ensure that the components will understand each other. Then we will be able to break down the problem and start more focused research projects to prototype each of the components.