Increased awareness of how to use AI as a tool for road safety by a new SAFER project
Classifying traffic safety on the state road network is a time-consuming and costly task for the Swedish Transport Administration. With today's collected amount of data, there is a need to investigate whether it is possible to identify objects that are currently linked to fixed objects in the roadside area. The initiative has the potential to be a support in the work of improving knowledge about the road side area as there is a lack of information about this today. Examples of fixed objects are game fence, posts, signs, trees, stones etc. By using machine learning technology, our hypothesis is that we can train an algorithm to automatically analyze and detect objects based on collected images and point clouds.
The new project “Recommendation for methodology on AI-based Road-side object identification” aim to investigate and specify the Swedish Transport Administration’s requirements for Road-side image processing, e.g. what is state of the art and what has already been done, are there existing solutions we can further utilize? The project team will also investigate and specify possible data scources, e.g. road scanning data and vehicle camera images from reference road areas. Processing possibilities will also be worked through e,g, requirements and costs for data storage and data access interfaces. Finally, the draft requirement for Proof of Concept for a road-side image processing project will be set up.
The result aim to be a recommendation for methods and requirements for data processing, including also how to set-up a Proof of Concept, for road-side object processing. The tool will help us to create very precise dynamic maps that can optimize road maintenance and information to automated vehicles in their self-driving mode.
The partners involved in this project are The Swedish transport administration, Chalmers, Asymptotic and Volvo Cars. SAFER’s research area Safety performance evaluation will host the project.