基于YOLOv5的井盖隐患智能识别研究  

Research on Intelligent Recognition of Hidden Dangers of Manhole Covers Based on YOLOv5

作　　者：黄健[1] 向思怡 HUANG Jian;XIANG Siyi(Xijing University,Xi'an 710123,China)

机构地区：[1]西京学院,陕西西安710123

出　　处：《现代信息科技》2025年第3期68-72,78,共6页Modern Information Technology

基　　金：陕西省大学生创新创业计划项目(S202412715035)。

摘　　要：作为城市地下管网系统的关键防护构件,井盖的安全状态直接影响市政设施运维效率与公共安全。针对传统井盖安全隐患检测方法存在效率低、误报率高等问题,文章提出基于改进YOLOv5的智能检测方法。通过构建多尺度特征融合机制,文章结合五折交叉验证法对标注数据集进行模型训练,实现了对破损、移位、缺失等典型安全隐患的精准识别。实验结果表明,改进模型在自建井盖数据集上的平均精度(mAP)达95.2%,相比于YOLOv4模型,精度和检测速度均有所提升。该算法通过优化特征金字塔结构与损失函数,有效增强了复杂路面场景下的目标表征能力,为城市基础设施智能化运维提供了可靠的技术支持。As a key protective component of the urban underground pipe network system,the safety status of manhole covers directly affects the operation and maintenance efficiency and public safety of municipal facilities.To address the inefficiency and high false alarm rate of traditional methods for detecting hidden dangers in manhole covers,this paper proposes an intelligent detection method based on improved YOLOv5.By constructing a multi-scale feature fusion mechanism,this paper combines the 5-fold cross-validation method to train the model of the labeled dataset,and realizes the accurate identification of typical safety hazards such as breakage,displacement,loss,and so on.The experimental results show that the mean Average Precision(mAP)of the improved model on the self-constructed manhole cover dataset is 95.2%.Compared with the YOLOv4 model,the accuracy and detection speed are improved.By optimizing the Feature Pyramid Network structure and loss function,the algorithm effectively enhances the target representation ability in complex road scenarios,providing reliable technical support for intelligent operation and maintenance of urban infrastructure.

关 键 词：井盖隐患 YOLOv5 目标检测 激活函数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP39[自动化与计算机技术—控制科学与工程]