Scholarlink

<p>Intersection safety continues to be a significant issue, with the potential for increased accidents involving vulnerable road users. This study proposes a novel approach to improve intersection safety through real-time video monitoring using the cutting-edge YOLOv9 object detection algorithm. The proposed system utilizes strategically placed cameras to capture live traffic footage, which undergoes image enhancement techniques before being processed by YOLOv9 for accurate detection and localization of VRUs. YOLOv9's architecture, GELAN (Generalized Efficient Layer Aggregation Network), and training approach, PGI (Programmable Gradient Information), enable efficient and precise object detection. The system is trained using a comprehensive dataset, and its performance is evaluated using standard metrics. The experimental results demonstrate an overall mAP (mean Average Precision) of 0.89 on the test set. The YOLOv9-based system outperforms other state-of-the-art object detection methods while maintaining real-time performance with a frame rate of 52 FPS. A simulation study using real-world traffic data demonstrates the system's potential to significantly improve intersection safety, resulting in a 62.5% reduction in potential VRU-vehicle conflicts and a 68% reduction in near-miss incidents. Compared to current approaches, the proposed system has capabilities that include high accuracy, real-time processing, and flexibility that can handle different intersection scenarios.</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

Real-time object detection and warning for road intersection safety: A YOLOv9-based approach

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Manish M Narkhede,
  • Rahul S Chaudhari,
  • Nilkanth B Chopade

摘要

Intersection safety continues to be a significant issue, with the potential for increased accidents involving vulnerable road users. This study proposes a novel approach to improve intersection safety through real-time video monitoring using the cutting-edge YOLOv9 object detection algorithm. The proposed system utilizes strategically placed cameras to capture live traffic footage, which undergoes image enhancement techniques before being processed by YOLOv9 for accurate detection and localization of VRUs. YOLOv9's architecture, GELAN (Generalized Efficient Layer Aggregation Network), and training approach, PGI (Programmable Gradient Information), enable efficient and precise object detection. The system is trained using a comprehensive dataset, and its performance is evaluated using standard metrics. The experimental results demonstrate an overall mAP (mean Average Precision) of 0.89 on the test set. The YOLOv9-based system outperforms other state-of-the-art object detection methods while maintaining real-time performance with a frame rate of 52 FPS. A simulation study using real-world traffic data demonstrates the system's potential to significantly improve intersection safety, resulting in a 62.5% reduction in potential VRU-vehicle conflicts and a 68% reduction in near-miss incidents. Compared to current approaches, the proposed system has capabilities that include high accuracy, real-time processing, and flexibility that can handle different intersection scenarios.