Scholarlink

<p>Ensuring the safety of vulnerable road users (VRUs), such as pedestrians, bicyclists, and E-scooter riders, at intersections is crucial for sustainable mobility. Many VRUs often disregard designated signals due to unclear button placement or the inconvenience of push-button systems, leading to crossing violations and reduced confidence in traffic infrastructure. This study introduces VRUCrossSafe, a real-time framework that utilizes computer vision and artificial intelligence to predict the crossing intentions of VRUs. By processing video feeds from cameras at signalized intersections, VRUs detection, tracking, and pose estimation are applied to extract the required features. The ensemble model is tested on 589 VRUs during daytime and nighttime, achieving crossing prediction accuracy of 94.67% while maintaining real-time processing at 33 frames per second. VRUCrossSafe enhances signal performance by automating pedestrian signal activation, ensuring adequate crossing time, and reducing crossing violations. VRUCrossSafe promotes sustainable transportation adoption while fostering safer and more efficient intersections.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

VRUCrossSafe for crossing intention prediction of vulnerable road users for improving safe crossing at intersections

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月21日

查看全文
  • Ahmed S. Abdelrahman,
  • Zubayer Islam,
  • Mohamed Abdel-Aty

摘要

Ensuring the safety of vulnerable road users (VRUs), such as pedestrians, bicyclists, and E-scooter riders, at intersections is crucial for sustainable mobility. Many VRUs often disregard designated signals due to unclear button placement or the inconvenience of push-button systems, leading to crossing violations and reduced confidence in traffic infrastructure. This study introduces VRUCrossSafe, a real-time framework that utilizes computer vision and artificial intelligence to predict the crossing intentions of VRUs. By processing video feeds from cameras at signalized intersections, VRUs detection, tracking, and pose estimation are applied to extract the required features. The ensemble model is tested on 589 VRUs during daytime and nighttime, achieving crossing prediction accuracy of 94.67% while maintaining real-time processing at 33 frames per second. VRUCrossSafe enhances signal performance by automating pedestrian signal activation, ensuring adequate crossing time, and reducing crossing violations. VRUCrossSafe promotes sustainable transportation adoption while fostering safer and more efficient intersections.