Evaluation of SIMMARC: An Audiovisual System for the Detection of Near-Miss Accidents

In this paper, we present and evaluate a system that automatically identifies hazardous traffic situations using visual and acoustic sensors. The system has been installed at three locations in Austria and several months of audio and video data have been analyzed. We evaluate the accuracy of the employed data analysis algorithms as well as the usefulness of the detected events for the overall task of assessing the risk potential of a road intersection. Our results show that the long-term analysis made possible by the proposed system leads to a better understanding of the risk potential of traffic areas, and can finally serve as a basis for defining and prioritizing improvements.