基于YOLOv8的塔式起重机司机疲劳作业监测算法  

Monitoring algorithm of tower crane driver’s fatigue work based on YOLOv8

作　　者：秦力 Qin Li

机构地区：[1]上海市浦东新区特种设备监督检验所,上海200062

出　　处：《起重运输机械》2025年第7期75-79,共5页Hoisting and Conveying Machinery

摘　　要：疲劳作业是塔式起重机司机常见且严重的一个安全隐患,可能导致重大事故和财产损失。针对这一问题,文中提出了一种基于YOLOv8的塔式起重机司机疲劳作业监测算法,通过对目标图像的检测与行为分析技术,实时监测司机的疲劳状态(如闭眼、打哈欠、低头等)。该方法结合YOLOv8的强大目标检测能力、MobileNetv2的轻量化分类优势,以及ECA模块对通道依赖的高效建模,显著降低计算复杂度,非常适合资源受限的嵌入式设备。实验表明,该模型在测试集上的mAP高达92.1%,优于未使用ECA模块的89.3%和基于SSD算法的76.8%。该算法为塔式起重机司机安全作业提供了一种高效、精准、可靠的解决方案,有助于提升建筑施工行业的安全水平。Fatigue operation poses a significant risk of major accidents and property losses,representing a common and severe safety hazard for tower crane drivers.To address this issue,a fatigue monitoring algorithm for tower crane drivers based on YOLOv8 is proposed.This algorithm can monitor drivers’fatigue states in real time,such as closing eyes,yawning,and bowing,by detecting target images and analyzing behaviors.The method integrates the robust target detection capabilities of YOLOv8,the lightweight classification advantages of MobileNetv2,and the efficient modeling of channel dependencies via the ECA module.This combination significantly reduces computational complexity,making it highly suitable for resource-limited embedded devices.Experimental results demonstrate that the model can achieve a mean Average Precision(mAP)of 92.1%on the test set,significantly higher than that of the algorithm without the ECA module(89.3%)and SSD-based algorithms(76.8%).This enhanced performance underscores the algorithm’s efficiency,accuracy,and reliability.By providing a robust solution for monitoring the safe operation of tower crane drivers,the algorithm contributes to elevating the overall safety in the construction industry.

关 键 词：塔式起重机 MobileNetv2 YOLOv8 疲劳作业 深度学习 

分 类 号：TH213.3[机械工程—机械制造及自动化] TP242[自动化与计算机技术—检测技术与自动化装置]