基于改进YOLOv8s的电动自行车驾乘人员头盔检测  

Helmet detection of electric bicycle riders based on improved YOLOv8s

作　　者：徐群子 徐国栋[1] 刘畅[1] XU Qunzi;XU Guodong;LIU Chang(School of Machinery and Transportation,Southwest Forestry University,Kunming 650224,China)

机构地区：[1]西南林业大学机械与交通学院,昆明650224

出　　处：《计算机应用文摘》2025年第2期61-63,共3页

摘　　要：为解决现有电动自行车驾乘人员头盔检测算法存在的漏检、误检问题,文章提出了一种改进的YOLOv8s算法模型。首先,向模型中添加了PSA注意力机制,使模型对目标特征的提取能力得到了增强。然后,使用Shape IoU损失函数进行优化,使边界框回归更加准确。最后,利用BiFPN来处理尺度变化和遮挡等复杂场景下的特征融合,从而提高模型的性能和鲁棒性。实验结果显示,在自建数据集上,改进后模型的mAP50比基础模型提高了0.4%,达到了91.6%,且其整体性能比其他算法模型更为优越。To address the issues of missed and false detections in existing helmet detection algorithms for electric bicycle riders,this article proposes an improved YOLOv8s algorithm model.Firstly,the PSA attention mechanism was added to the model,enhancing its ability to extract target features.Then,optimize the bounding box regression using the Shape IoU loss function to make it more accurate.Finally,BiFPN is utilized to handle feature fusion in complex scenes such as scale changes and occlusions,thereby improving the performance and robustness of the model.The experimental results showed that on the self built dataset,the mAP50 of the improved model increased by 0.4%compared to the base model,reaching 91.6%,and its overall performance was superior to other algorithm models.

关 键 词：头盔检测 注意力机制 损失函数 特征金字塔网络 

分 类 号：TP389[自动化与计算机技术—计算机系统结构]