GAIA: AI for capturing of geographical base data

Research Project, 2023 – 2025

The demand for geographic data is changing as the complexity of the development of cities and communities increases.To meet the challenges of a sustainable urban development as well as climate adaption there is a demand for higher quality, completeness and actuality of geodata.
The demand for geographic data is changing as the complexity of the development of cities and communities increases. To meet the challenges of a sustainable urban development as well as climate adaption there is a demand for higher quality, completeness and actuality of geodata. Moreover, there is also a need for classification and descriptive data for the mapped objects.  Geodata provided by municipalities is one of the corner stones of the process of developing the built environment. Municipal basemaps forms the basis of strategic documents such as urban development master plans, detail zoning plans, climate analyses and ecological analyses. In order to better support these applications, many municipalities are starting to produce digital twins for planning and management. In the not too distant future the 3D objects of the digital twins, rather than a 2D map such as the basemap, will form the basis for developing the built environment. However, with that follows massively increased amounts of data.  For the most part, these data are still being produced manually – a time consuming and relatively inefficient process. A large city with an active maintenance of basic geodata spends several thousand working hours each year on this.  By using machine learning, a sort of artificial intelligence (AI), to interpret and vectorize high resolution aerial imagery, GAIA attempts to automate the most time consuming aspects of mapping cities and municipalities.  All 290 municipalities in Sweden could replicate a method of this kind, as well as other authorities and private companies operating in mapping and geodata collection in Sweden. The method - along with more frequent collection of imagery, for example using UAV:s (Unmanned Aerial Vehicles) - thereby has the potential of saving vast amounts of work while significantly increasing the actuality and completeness of geodata.  The purpose of this project is to facilitate production and updating basic geodata by utilising AI for automation. The data produced forms the basis of digital twins as well as basemaps. In order to acheive the purpose, there are four quantifiable milestones. These set the ambition of accuracy in the results as well as the streamlining of the mapping process by at least 75%. The project is organized into six different parts: 1) Survey of current processes, 2) inventory and data collection, 3) tagging and processing data for machine learning, 4) developing task-specific algorithms, 5) automating the process, and 6) documenting and spreading information and insights.  The composition of the project group is optimized to ensure that the purpose and objectives of the project can be met in the best possible way. The five participating municipalities having the needs and acting as recipients ensure that the results are useful for municipalities on the whole: small, medium-sized and big municipalities. The representativs from the private and the academic sector are leading in their respective areas and provide expertise for the project work as well as establishing a ground for maintaining the results after the closing of the project.