Machine learning as a mapping tool: production of ground cover data

Important results from the project

The project aims to extract more qualitative data in the form of ground cover data based on existing orthophotos and elevation data. The project will be a learning project where knowledge and lessons are passed on within the organization to hopefully find additional uses for this technology. The project contributes to the follow-up of the municipality´s multi-year strategy of developing digital solutions, as for example, through the use of artificial intelligence (AI) to reduce the administrative work and devote more time to qualified tasks.

Expected long term effects

Within this project vectorized geodata has been produced from orthophotos and elevation data. By using machine learning as a method, the project wants to create a model that, like the human eye does, can distinguish and classify objects in an image. Additionally, the model can be developed further to identify other types of data and serve as a basis for future mappings. This way, updating geodata will become a more efficient and uniform process. In order to benefit other organizations, the model is based on open source code and can be shared.

Approach and implementation

During the implementation of the project, the group has used an agile working methodology. It involved planning, implementing and getting feedback from shorter sprints with defined constraints. This method has helped the group to become more compliant and adaptable as the project was progessing. Further, the method suits well for this learning project, which involves new technological aspects for the group. The project has been carried out according to the Samhällsbyggnadsförvaltningens project model and according to the terms set by Vinnova for granting funding.

External links

Projektet samt resultat i sin helhet presenteras via en kartberättelse, StoryMaps (ArcGIS Online)