基于改进YOLOv10的轻量化交通流量检测  

Lightweight Traffic Flow Detection Based on Improved YOLOv10

作　　者：白云 刘丹丹 BAI Yun;LIU Dandan(School of Aeronautics,Inner Mongolia University of Technology,Hohhot 010051,China)

机构地区：[1]内蒙古工业大学航空学院,内蒙古呼和浩特010051

出　　处：《现代信息科技》2025年第8期77-82,共6页Modern Information Technology

基　　金：2023自治区高校基本科研业务费项目-复杂交通场景下的环境感知关键技术研究(RZ2300001569);2023自治区科技计划-交通安全风险监测体系与预警模型关键技术研究(RZ2300002822)。

摘　　要：为解决传统交通流量统计算法在实时性、稳定性和精确度方面的不足,提出一种基于改进YOLOv10算法的目标检测算法。将改进后的YOLOv10算法与DeepSORT追踪技术相结合,构建新的交通流量统计架构。采集实际路况的交通场景视频,对该框架的准确性进行验证。实验结果表明,相较于原算法,改进后的新算法在车辆检测方面平均精度提升了1%,视频车流量统计精度提升了3.46%。In order to solve the shortcomings of traditional traffic flow statistics algorithms in real-time,stability and accuracy,a target detection algorithm based on improved YOLOv10 algorithm is proposed.The improved YOLOv10 algorithm is combined with DeepSORT tracking technology to construct a new trafficflow statistical architecture.The traffic scene video of the actual road condition is collected to verify the accuracy of the framework.The experimental results show that compared with the original algorithm,the average accuracy of the improved new algorithm in vehicle detection is improved by 1%,and the statistical accuracy of video trafficflow is improved by 3.46%.

关 键 词：计算机视觉 交通流量统计 YOLO DeepSORT 智能交通 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]