Scholarlink

Existing vision-based traffic load monitoring systems only provide vehicle bounding boxes, which makes it challenging to accurately determine the load positions. This paper proposes a framework to determine real-time traffic load distribution. The proposed framework uses the YOLO-v5 model to identify vehicles on the monitoring screen and makes reasonable assumptions about the geometric shape of the vehicles. These assumptions are used as prior information to propose an indirect calculation method for vehicle equivalent concentrated load, and a load position coordinates dataset is labeled. Finally, the statistical law of the labeled data is analyzed to verify its normality, and the maximum likelihood estimation of the load coordinate is determined. The proposed method is tested in the field, combined with a Kalman filter to track vehicles and optimize the load trajectory. The results show that the proposed method significantly improves the accuracy of the original algorithm and has strong practical application value.

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 Traffic Load Monitoring Framework Based on Deep Learning Model and Statistical Regularities of Vehicle Shape Prior Information

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Boqiang Xu,
  • Zhijun Shi,
  • Chao Liu

摘要

Existing vision-based traffic load monitoring systems only provide vehicle bounding boxes, which makes it challenging to accurately determine the load positions. This paper proposes a framework to determine real-time traffic load distribution. The proposed framework uses the YOLO-v5 model to identify vehicles on the monitoring screen and makes reasonable assumptions about the geometric shape of the vehicles. These assumptions are used as prior information to propose an indirect calculation method for vehicle equivalent concentrated load, and a load position coordinates dataset is labeled. Finally, the statistical law of the labeled data is analyzed to verify its normality, and the maximum likelihood estimation of the load coordinate is determined. The proposed method is tested in the field, combined with a Kalman filter to track vehicles and optimize the load trajectory. The results show that the proposed method significantly improves the accuracy of the original algorithm and has strong practical application value.