基于深度学习的安全帽智能识别系统设计与实现  

Study on the Design and Implementation of an Intelligent Identification System for Safety Helmets Based on Deep Learning

作　　者：郭普特 郑斌[1] 黄敏[1] 苏洁 李铖杰 韦天健 刘宇 Guo Pute;Zheng Bin;Huang Min;Su Jie;Li Chengjie;Wei Tianjian;Liu Yu(School of Computer and Communication Engineering,Changsha University of Science and Technology,Changsha 410114)

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410114

出　　处：《中阿科技论坛（中英文）》2022年第4期137-141,共5页China-Arab States Science and Technology Forum

基　　金：2020年湖南省大学生创新创业训练项目(S202113635008)。

摘　　要：近年来,随着深度学习算法进入机器视觉领域,计算机视觉检测技术得到了进一步的发展,基于深度学习的检测技术能够完成快速、精确、稳定的目标检测。本文针对建筑制造业的作业现场存在工人不佩戴安全帽的问题,对现有的深度学习目标检测模型YOLOv3进行改进,采用PyQt5开源框架设计了跨平台的终端用户交互界面,开发了能够对作业人员安全帽佩戴情况进行实时准确检测的识别算法。经试验,最终系统检测速度可达20FPS以上,模型mAP可达86.7%,基本实现安全帽检测系统的智能化,可一定程度上填补建筑制造行业中安全帽智能检测系统市场的空缺。In recent years,with the entry of deep learning algorithms into the field of machine vision,computer vision detection technology has enjoyed rapid development.Detection technology based on deep learning can realize fast,accurate and stable target detection.In view of the problem of workers not wearing safety helmets in the construction industry,by perfecting the existing deep learning target detection model YOLOv3,a cross-platform end-user interface is designed through the PyQt5 open source framework,an identification algorithm for real-time and accurate detection of helmet wearing is established.Testing results have shown that the system can basically realize the intelligence of the safety helmet detection with the detection speed of more than 20FPS,in addition,the speed of the mAP model can reach 86.7%,which can bridge the gap in the safety helmet intelligent detection system market in the construction industry to a certain extent.

关 键 词：深度学习 YOLOv3 PyQt5 安全帽 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论] TP391.41[自动化与计算机技术—计算机科学与技术]