Event-Based Safety Indicator Analysis of a Surround View Monitoring (SVM)-Based Pedestrian Detection System Using Real-World Urban Data

Pedestrian safety in urban areas remains a critical concern, especially for vehicles with large blind spots. This study examined associations between surround view monitoring (SVM)-defined risk zones and safety-related indicators using real-world driving data from school zones and pedestrian crash-prone areas in Seoul, South Korea. Vehicle on-board diagnostics, RTK positioning, SVM object-detection outputs, and video data were collected, and vehicle–vulnerable road user interactions were reconstructed as events. Detection-to-braking time (DBT), time to collision (TTC), time-exposed TTC (TET), time-integrated TTC (TIT), TIT/TET, and jerk were used to compare danger and warning zones. The danger zone showed lower estimated TTC and higher TIT/TET values than the warning zone, indicating shorter time margins and higher average risk intensity during below-threshold TTC intervals. The warning zone showed higher longitudinal jerk, suggesting repeated vehicle-control adjustments during longer events. In the four-day repeated-driving dataset, the 95th percentile total jerk in the danger zone decreased over time; however, this uncontrolled trend was interpreted as exploratory rather than causal. These findings suggest that SVM-defined risk zones are associated with differences in event-level risk and vehicle dynamic indicators, supporting their potential use in event-based safety assessment.