As the national mapping agency, one of our main tasks is to keep our mapping and spatial data as up to date as possible. To ensure this, we need to know when new houses have been built, when sites have been re-developed, and when any other changes have occurred which will affect our mapping. We obtain information about change from many sources, including local authorities, surveyors on the ground, and from aerial photography. Detecting the changes from photos is usually a laborious and time-consuming manual process, in which a team of people compare the aerial images with the features in our mapping, but we would like to make this a more automatic process.
Extracting change information from aerial photography
Our change detection research team has investigated several methods of detecting change and is now concentrating on the most promising - using eCognition® object-based classification techniques. This process uses digital aerial photography, and the height data derived from it via photogrammetry, to identify the features we are most interested in. These are then compared with the same features in our OS MasterMap® database, to identify the new features, the features which are no longer present, and any other significant differences between the two datasets. We are confident that we can now automatically find changes such as new and demolished buildings, to a degree of accuracy similar to that of the manual process. Our challenge is to extend the techniques to find changes to vegetation, field boundaries, roads and water features.