基于YOLOv4-tiny的安全帽检测算法研究  

Research on helmet detection algorithm based on YOLOv4-tiny

作　　者：赵建光[1] 韩泽山 范晶晶[1] 张君秋 ZHAO Jianguang;HAN Zeshan;FAN Jingjing;ZHANG Junqiu(Hebei University of Architecture,Zhangjiakou,Hebei 075000)

机构地区：[1]河北建筑工程学院,河北张家口075000

出　　处：《河北建筑工程学院学报》2023年第4期240-245,共6页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：基于YOLOv5s算法的吸烟行为检测研究(XY2023024)。

摘　　要：为了可以有效监管施工人员佩戴安全帽的情况,使用YOLOv4-tiny目标检测算法进行佩戴安全帽检测。YOLOv4-tiny模型在维持精确度的同时,计算量更少,检测速度更快,更适用于实时的安全帽佩戴检测。但是YOLOv4-tiny在复杂的工作场景中容易出现错检或漏检情况。为解决这一问题,在原模型的基础上添加CBAM、SE、ECA、CA注意力机制,将几种不同的注意力机制进行比较,解决模型检测效果差的问题,使用改进的YOLOv4-tiny算法可以更好地完成安全帽检测任务。In order to effectively supervise the wearing of safety helmets by construction workers,the YOLOv4-tiny target detection algorithm is used to detect safety helmets.While maintaining accuracy,the YOLOv4-tiny model requires less computation and faster detection,making it more suitable for real-time helmet wearing detection.However,YOLOv4-tiny is prone to misdetection or missed detection in complex work scenarios.To solve this problem,CBAM,SE,ECA,and CA attention mechanisms are added on the basis of the original model,and several different attention mechanisms are added.For comparison,to solve the problem of poor model detection effect,the improved YOLOv4-tiny algorithm can better complete the helmet detection task.

关 键 词：安全帽检测 YOLOv4-tiny 注意力机制 检测效果 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]