Prediction of Pedestrian Crossing Intentions at Intersections Based on Long Short-Term Memory Recurrent Neural Network
Shile Zhang, Mohamed Abdel‐Aty, Jinghui Yuan, Pei Li
Abstract
Traffic violations of pedestrians at intersections are major causes of road crashes involving pedestrians, especially red-light crossing behaviors. To predict the pedestrians’ red-light crossing intentions, video data from real traffic scenes are collected. Using detection and tracking techniques in computer vision, some pedestrians’ characteristics, including location information, are generated. A long short-term memory neural network is established and trained to predict pedestrians’ red-light crossing intentions. The experimental results show that the model has an accuracy rate of 91.6% based on internal testing at one signalized crosswalk. This model can be further implemented in the vehicle-to-infrastructure communication environment and prevent crashes because of the pedestrians’ red-light crossing behaviors.