Optimal Sequential Detection of Freeway Accidents

Accurate traffic accident detection constitutes a key step in a variety of processes ranging from improving road safety conditions and route navigation, to making informed decisions in urban planning. This paper proposes a Bayesian quickest change detection framework for accurate detection in near-real-time of freeway accidents based on speed sensors. To exploit the spatial correlation between sensors, and to improve the robustness of the proposed solution, four aggregation schemes are also proposed. The effectiveness of the proposed approach is illustrated on a real-world dataset from the I405 freeway in the Los Angeles County. Comparison with the state-of-the-art demonstrates the superiority of the proposed approach.