摘要
公共场所自动扶梯的人员跌倒问题时有发生,为保障乘客安全,本文提出一种基于YOLOv5s的实时检测算法。该算法在YOLOv5s网络结构的基础上进行改进,利用GhostNet模块重构Backbone部分以显著减少特征图的冗余计算,同时在Backbone中引入CBAM(Convolutional Block Attention Module,注意力机制),以增强特征提取能力。相比原始网络,该算法在保持较高检测精度的同时,实现了更优的轻量化设计,其检测准确率保持在87.6%,召回率保持在85.3%,为公共场所自动扶梯的安全监测提供了一种高效可靠的解决方案。
The problem of people falling down on escalators in public places occurs from time to time.In order to ensure the safety of passengers,this paper proposes a real-time detection algorithm based on YOLOv5s.The algorithm is improved on the basis of YOLOv5s network structure.The Backbone part is reconstructed by GhostNet module to significantly reduce the redundant calculation of feature graph.At the same time,convolutional block attention module(CBAM)is introduced into the Backbone to enhance the ability of feature extraction.Compared with the original network,the algorithm not only maintains high detection accuracy,but also achieves better lightweight design.Its detection accuracy rate is maintained at 87.6%,and the recall rate is maintained at 85.3%.It provides an efficient and reliable solution for the safety monitoring of escalators in public places.
作者
夏小松
唐川
赵再友
马永军
刘传奇
XIA Xiaosong;TANG Chuan;ZHAO Zaiyou;MA Yongjun;LIU Chuanqi(Chongqing Special Equipment Inspection and Research Institute,Chongqing,401121;Key Laboratory of State Administration for Market Regulation(Safety of Mechanical and Electrical Equipment in Western Complex Environments),Chongqing,401121)
出处
《中国特种设备安全》
2025年第S1期29-33,38,共6页
China Special Equipment Safety
基金
重庆市科研机构绩效激励引导专项项目(CSTB2023JXJL-YFX0007)。
