Pedestrian-Vehicle Collision Avoidance Support System Using Multiple Pedestrian Motion

In order to prevent traffic accidents between pedestrians and vehicles, we are researching a Pedestrian – Vehicle Collision Avoidance Support System (P-VCASS). We propose a new system that can handle multiple pedestrians by judging their motion by machine learning of random forest. Each pedestrian terminal judges his motion from acceleration and angular velocity sensors and transmits the judgment results and GPS information to the vehicle’s terminal. To verify the effectiveness of the proposed system, we conducted some experiments in actual situation. The results showed that it was possible to warn the driver with 93% reliability.