Inference Method for Abnormal Traffic Events in Highway Monitoring Blind Spots

Highways have the characteristic of closed-loop traffic flow, presenting unique challenges for traffic surveillance and management. This study focuses on addressing the problem of detecting abnormal events in areas that cannot be directly monitored. These areas typically lack the necessary infrastructure, such as cameras, radar, or other sensing devices, making it difficult to identify traffic abnormal events such as accidents or congestion. To tackle this challenge, this paper proposes a discriminative algorithm for detecting abnormal events in highway surveillance blind spots. Firstly, we simulate traffic scenarios including both abnormal events and normal situations using the VISSIM software, and select speed, density, and occupancy as feature parameters. Subsequently, this study applies the K-means clustering algorithm to judge whether an abnormal event occurs. Experimental results show that the proposed algorithm exhibits high precision (91.4%) in identifying abnormal events, with a false negative rate of only 4.17%. Moreover, the algorithm demonstrates good robustness against disturbances when individual raw parameters become anomalous.