Publication

This study aimed to determine whether perceived risk, as rated by a large group of human participants, could be predicted from basic features extracted using computer vision. To achieve this, we developed a framework for estimating perceived traffic risk in various traffic scenes, encompassing diverse scenarios. By employing computer vision techniques, we quantified the information in the images, such as the number of vehicles and their distances. Using a random forest regression model, we built a prediction model to assess risk perception, aiming to distinguish safety between different traffic scenarios.