Real-time crash prediction based on high definition monitoring systems

This paper presents an innovative approach to investigate the inner mechanism between traffic status and crash potential based on High Definition Monitoring Systems (HDMS) data. HDMS records delicate vehicle trajectory data and characteristic details. Matched case-control method and Support Vector Machines (SVMs) were employed to identify risk status. The grid search method was utilized to find the optimized parameters for the decision function of the SVMs. The results indicate that the SVMs classifier could successfully classify 91.49% of the crashes on the test dataset with a false alarming rate 2.63%. This study contributes to decision supports for road safety management administrations.