Scholarlink

High-quality vehicle trajectory data sets with less error are becoming increasingly important for the development of autonomous driving systems. With the widespread use of drones and radar vehicle detectors in intelligent transportation, this paper proposes two vehicle trajectory extraction techniques for these two devices, aiming to enhance the diversity of vehicle recognition and speed measurement methods. Firstly, for the scenario of a drone hovering and capturing footage from high altitudes, an algorithm model based on YOLOv8 and DeepSORT is designed, and a manually labeled dataset is used to train the detector. Secondly, redevelopment was conducted on the Hikvision radar vehicle detector, enabling it to perform full-range vehicle trajectory uploading for a 120-meter road segment. Finally, an accuracy and error analysis of the two proposed methods was performed. Our code is now open source and available for download at https://github.com/Dexter753/ . \(\dots \)

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

Comparison of Vehicle Trajectory Data Collection Techniques Using Machine Vision and Radar

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhen Yan,
  • Shiwen Du,
  • Yangming Xing,
  • Lin Li,
  • Zhuo Wang,
  • Hongwei Song

摘要

High-quality vehicle trajectory data sets with less error are becoming increasingly important for the development of autonomous driving systems. With the widespread use of drones and radar vehicle detectors in intelligent transportation, this paper proposes two vehicle trajectory extraction techniques for these two devices, aiming to enhance the diversity of vehicle recognition and speed measurement methods. Firstly, for the scenario of a drone hovering and capturing footage from high altitudes, an algorithm model based on YOLOv8 and DeepSORT is designed, and a manually labeled dataset is used to train the detector. Secondly, redevelopment was conducted on the Hikvision radar vehicle detector, enabling it to perform full-range vehicle trajectory uploading for a 120-meter road segment. Finally, an accuracy and error analysis of the two proposed methods was performed. Our code is now open source and available for download at https://github.com/Dexter753/ . \(\dots \)