Vision-Based Personal Safety Messages (PSMs) Generation for Connected Vehicles

Abstract

Vehicle-to-pedestrian (V2P) communication can significantly improve pedestrian safety in a connected vehicle environment. However, pedestrian safety is hindered as pedestrians often do not carry hand-held devices that provide low latency wireless communication (e.g., dedicated short-range communication (DSRC)-enabled device or emerging 5G-enabled cell phone) to communicate with nearby connected vehicles. The contribution of this paper lies in developing a vision-based approach to generate personal safety messages (PSMs) in real-time utilizing video streams from roadside traffic cameras, following the standard of the Society of Automotive Engineers (SAE) (SAE J2945) that can be used by connected vehicle pedestrian safety applications. Our analysis reveals that the vision-based approach can estimate pedestrians' location and speed more accurately than existing DSRC-enabled pedestrian hand-held devices. A system-level validation was performed by developing a connected vehicle related pedestrian safety application, “pedestrian in signalized crosswalk warning (PSCW),” that uses the generated PSMs from our vision-based approach. The results from the calculated average time-to-collision (TTC) value demonstrate the efficacy of our method in generating real-time collision warnings to avoid possible vehicle-pedestrian collisions. Our analysis also shows that the vision-based pedestrian safety warning system satisfies the latency requirement for the PSCW safety application in a connected vehicle environment.