In KI4Safety, AI methods are applied to big data to identify safety-relevant patterns and analyse the effect of traffic management features on accidents, which supports road safety work.
Deep learning recognises safety-relevant features in road traffic - a look at the geodatabase.
In practice, it is becoming increasingly difficult to reduce the number of road accidents. Innovative approaches are needed to derive effective and cost-efficient measures. The factors influencing human behaviour and the occurrence of traffic accidents are very diverse. The relevant data sources and possible combinations are just as diverse, and it takes a long time to analyse them comprehensively and comparatively.
Project objective
In KI4Safety, artificial intelligence methods are applied to big data in order to improve the evidence regarding the effect of traffic management features on the occurrence of accidents. In particular, safety-relevant infrastructure features and patterns are automatically recognised in a large amount of aerial and traffic image data and used for impact analysis. The catalogue of measures against accident clusters is used to demonstrate that AI can be used to gain insights from big data that conventional investigations have so far only provided with great effort or not at all. The system offers approaches to support practical road safety and planning work.
Realisation
Data from relevant factors influencing the occurrence of accidents, such as infrastructure data, the type of traffic routing, traffic engineering parameters and traffic control data are merged with corresponding accident data. With the help of artificial intelligence methods, it is possible to handle large samples in comparative analyses, which improves the situation in terms of statistical significance. The system is being tested and optimised in close cooperation with traffic planners, the police and road safety researchers from various institutions.
Recognising safety-relevant features of road traffic management in generally available Google Street View traffic images.
