基于改进YOLOv5和Deep SORT的桥梁车辆识别及跟踪研究
Study of Vehicle Detection and Tracking on Bridge by Improved YOLOv5 and Deep SORT
赵智勇1
作者信息
- 1. 中交第一公路勘察设计研究院有限公司,陕西西安 710075
- 折叠
摘要
掌握真实的车辆荷载情况对桥梁设计及智能管养具有重要意义.为此,基于计算机视觉技术和深度学习,建立了-种用于桥梁上多车检测和跟踪的算法.首先,收集交通监控视频建立了多种类型车辆的外观特征数据库.其次,建立了多车检测算法,并在所搭建的数据库上对其进行训练和测试.随后,将性能最佳的检测算法与跟踪算法相结合,进而完成桥梁上多车目标的连续跟踪.最后,依托某跨海大桥的交通监控视频对所提方法进行了验证,并评估了算法的可靠性和准确性.实验结果表明:提出的多车检测和跟踪算法的检测准确率较高,跟踪效果较好,在视频序列中稳定性较好,可成功完成桥梁上多车的连续跟踪任务.研究成果可为后续桥梁设计及智能化管养提供数据参考.
Abstract
It is of great significance to know the real vehicle load condition for bridge design and intelligent mainte-nance.Therefore,based on computer vision technology and deep learning,the multi-vehicle detection and tracking algorithm on the bridge is established in this paper.Firstly,a vehicle appearance dataset containing multiple types is established by collecting traffic surveillance videos.Secondly,the multi-vehicles detection algorithms is estab-lished and trained and tested on the dataset.Then,the algorithm with the best performance is combined with the best tracking algorithm to complete the multi-vehicle target tracking on the bridge.Finally,based on the traffic monitoring data of a long-span bridge,the improved effect of the algorithm is verified,and the reliability and accu-racy of the proposed algorithm are verified.The experimental results show that the proposed multi-vehicle detection and tracking algorithm has high detection accuracy,better tracking effect and stability in video sequences,which can successfully complete the continuous tracking of multi-vehicles on bridges.The research results can provide da-ta reference for the subsequent intelligent management and maintenance of bridges.
关键词
桥梁工程/计算机视觉/车辆荷载/目标检测/目标跟踪Key words
bridge engineering/computer vision/vehicle load/target detection/target tracking引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据