基于标签重匹配及边缘端部署的轻量化交通量统计方法  

A lightweight traffic counting method based on vehicle label rematching and edge-end deployment

作　　者：张森 刘峻池 王林飞 齐冲冲 阿西平 陶大鹏[2] ZHANG Sen;LIU Junchi;WANG Linfei;QI Chongchong;A Xiping;TAO Dapeng(Yunnan Provincial Highway Science and Technology Research Institute,Kunming 650051,China;School of Information Science and Engineering,Yunnan University,Kunming 650504,China;Yunnan United Vision Technology Co.,Ltd.,Kunming 650504,China;Kunming Highway Administration Bureau,Kunming 650220,China)

机构地区：[1]云南省公路科学技术研究院,云南昆明650051 [2]云南大学信息学院,云南昆明650504 [3]云南联合视觉科技有限公司,云南昆明650504 [4]昆明公路局,云南昆明650220

出　　处：《应用科技》2024年第5期91-100,共10页Applied Science and Technology

基　　金：云南省交通运输厅项目(2022-23-3)。

摘　　要：为解决低等级普通道路交通观测站在开展自动化交通量统计时面临的车型间识别混淆、边端计算资源和大规模数据传输受限等问题,在YOLOv8和DeepSORT模型的基础上设计了一种类别标签重匹配策略,可以在目标遮挡、远距离、视角不佳等场景下实现高质量的九型车识别与统计。其次,设计一种基于Jetson NX嵌入式平台的边缘端模型部署和一种基于TensorRT模型优化加速方法,用于九型车流量统计模型的现场部署和边端推理。实验证明所提方法在实际道路下可达到97.9%的统计精度和17.27 f/s的推理帧率,可实现在边端推理模式下速度和精度的良好平衡,能为道路交通调查提供精准、实时的车流量信息。To address the issues of vehicle classification confusion,limited edge computing resources,and large-scale data transmission constraints encountered by low-grade road traffic observation stations in automated traffic volume statistics,this paper proposes a category label re-matching strategy based on the YOLOv8 and DeepSORT models.This strategy enables high-quality recognition and counting of nine types of vehicles,even in challenging scenarios such as object occlusion,long distances,and unfavorable viewing angles.Additionally,an edge deployment solution based on the Jetson NX embedded platform,coupled with a TensorRT model optimization acceleration method,is designed for the on-site deployment and edge inference of nine types of vehicles traffic counting model.Experimental results demonstrate that the proposed method achieves a statistical accuracy of 97.9%and an inference speed of 17.27 f/s under real road conditions,striking a good balance between speed and accuracy in edge inference mode.This provides precise,real-time traffic flow information for road traffic surveys.

关 键 词：交通调查 九型车识别 交通量统计 目标检测 目标跟踪 边缘计算 YOLOv8 Jetson NX 

分 类 号：U495[交通运输工程—交通运输规划与管理]