Behavioral analysis of pedestrians using AI

During the project, traffic data was collected from a total of approximately 500 hours of video recordings from 32 different points in and around downtown Bergen. This was done using innovative tools for analyzing video recordings using AI. These tools utilize image recognition to identify all road users in a video recording and automatically classify them according to their road use group. Their movement patterns are also mapped and stored. In this project, a total of approximately 680,000 traffic movements were recorded. The collected data was further used to conduct various analysis of behavior for pedestrians.

One main issue in the assignment was to assess whether reduced sidewalk width due to outdoor dining could lead to poorer accessibility and reduced traffic safety for pedestrians.

The behavior analyses revealed several instances of crowding and reduced accessibility related to outdoor dining, with potentially increased risk for pedestrians and cyclists. However, the results indicate that this is primarily a concern on days with very high pedestrian volumes.

There were also cases where the establishment of outdoor dining affected pedestrian behavior but did not result in crowding. Nevertheless, it was found that this could pose challenges for individuals with strollers or users of wheelchairs and walkers.

As part of the delivery, a data-driven insight tool was also developed to analyze the collected data. This provides increased insight and added value for the client.