基于深度学习的校园交通监管系统  

Campus traffic supervision system based on deep learning

作　　者：张书豪 彭睿颖 高胜炜 田文冰 袁博宇 刘玮玮 ZHANG Shuhao;PENG Ruiying;GAO Shengwei;TIAN Wenbing;YUAN Boyu;LIU Weiwei(JSNU SPBPU Institute of Engineering,Sino-Russian Institute,Jiangsu Normal University,Xuzhou 221000,China;Taizhou Branch,China Telecom Corporation Limited,Taizhou 225300,China;School of Smart Education,Jiangsu Normal University,Xuzhou 221000,China)

机构地区：[1]江苏师范大学江苏圣理工学院-中俄学院,江苏徐州221000 [2]中国电信股份有限公司泰州市分公司,江苏泰州225300 [3]江苏师范大学智慧教育学院,江苏徐州221000

出　　处：《山东理工大学学报（自然科学版）》2025年第2期16-21,共6页Journal of Shandong University of Technology:Natural Science Edition

基　　金：2022年江苏省大学生创新创业训练计划项目(202210320096Y)。

摘　　要：针对校园内机动车辆管理混乱和违章行为频发的问题,设计了一个基于YOLOv5和CRNN模型的混合式车辆监控系统。该系统利用YOLOv5目标检测算法和CRNN车牌识别技术,能有效识别超速和违停等违章行为并自动发出预警,实现了对校园车辆的精确检测与管理。在不同校园场景下进行的实验结果表明,本文模型的车辆检测精准率较YOLOv3提高了3.39%,召回率和mAP较其他几种模型也均有显著提升。该监控系统在校园环境中具有较高的应用价值,不仅能够提升校园交通的安全性和管理效率,也能为智能交通系统的发展提供新的技术支持和实践案例。This paper addresses the prevalent issues of chaotic vehicle management and frequent traffic violations on campus by designing and implementing a hybrid vehicle monitoring system based on YOLOv5 and CRNN models.Utilizing YOLOv5 for object detection and CRNN for license plate recognition,the system achieves precise detection and management of campus vehicles,effectively identifying violations such as speeding and illegal parking and issues timely alerts.Experimental validation across different campus scenarios demonstrates a 3.39%enhancement in vehicle detection accuracy compared to YOLOv3,with significant improvements in recall rate and mAP compared to several other models.These results indicate that this monitoring system has the strong applicability in campus environments,not only enhancing traffic safety and management efficiency but also providing new technical support and practical examples for the development of intelligent transportation systems.

关 键 词：校园车辆监管系统 车辆识别 YOLOv5 CRNN 交通违章检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]