改进YOLOv5s算法的头盔佩戴实时检测系统  

A Real-time Helmet Wearing Detection System based on Deep Learning-improved YOLOv5s Algorithm

作　　者：毛雅 余锦 李倩南 吴青周 MAO Ya;YU Jin;LI Qiannan;WU Qingzhou(Huangshan University,Huangshan,China,245000)

机构地区：[1]黄山学院,安徽黄山245000

出　　处：《福建电脑》2025年第3期24-29,共6页Journal of Fujian Computer

基　　金：黄山学院2023年度大学生创新创业训练计划“基于深度学习——改进YOLOv5s算法的头盔佩戴实时检测系统”(No.202310375010);安徽省创新型省份建设补助资金专项资助项目(No.2020xzx004);安徽省高等学校科学研究重点项目(No.2023AH051371)资助。

摘　　要：为解决电动车头盔检测中因天气、角度等因素造成的检测效率不高的问题,本文提出一种基于改进YOLOv5算法的头盔佩戴实时检测系统。系统设计集成了高清摄像头的视频流输入、高效的算法处理以及实时检测结果输出。通过采用数据增强、结构优化和损失函数调整的技术手段,系统显著提升了在复杂环境下的识别精度和速度。测试结果表明,优化后的YOLOv5s算法头盔检测系统的检测准确率比原YOLOv5算法提升了1.1%。To solve the problem of low detection efficiency caused by factors such as weather and angle in electric vehicle helmet detection,this paper proposes a helmet wearing real-time detection system based on an improved YOLOv5 algorithm.The system design integrates high-definition camera video stream input,efficient algorithm processing,and real-time detection result output.By adopting techniques such as data augmentation,structural optimization,and loss function adjustment,the system significantly improves recognition accuracy and speed in complex environments.The test results show that the detection accuracy of the optimized YOLOv5s algorithm helmet detection system has improved by 1.1 percentage points compared to the original YOLOv5 algorithm.

关 键 词：头盔佩戴检测 电动车安全 YOLOv5算法 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]