基于改进YOLO v5的电网工人防具检测  

Grid Worker Armor Detection Based on Improved YOLO v5

作　　者：戴云周 路红[1] 纪陈阳 DAI Yunzhou;LU Hong;LU Huiwen(School of Automation,Nanjing Institute of Technology,Nanjing 211167,China;Linkage Rugao Software Co.,Ltd.,Rugao 226599,China)

机构地区：[1]南京工程学院自动化学院,江苏南京211167 [2]凌志软件如皋有限公司,江苏如皋226599

出　　处：《南京工程学院学报（自然科学版）》2023年第1期33-38,共6页Journal of Nanjing Institute of Technology(Natural Science Edition)

基　　金：南京工程学院研究生大学生科技创新项目(TB202217006)。

摘　　要：为了防止电网工人在高空作业时防具穿戴不当而导致安全事故发生,基于PyTorch框架提出一种改进型的轻量级防具检测算法.首先将YOLO v5s的Backbone用轻量级网络GhostNet来替换,降低网络复杂程度,减少参数量;然后针对工人在高空作业中的目标遮挡问题,通过改进随机擦除与Mosaic数据增强的方法模拟高空中防具被遮挡的情况;最后优化loss函数,采用Distance IoU替换常用的IoU,解决网络检测框位置偏差大的问题.采用平均精度均值mAP与模型大小来评估本文改进算法与其他算法对防具检测的效果,试验结果表明了本文改进方法的有效性,满足在移动端部署的需求.In order to prevent safety accidents caused by improper wearing of protective equipment when grid workers perform high-altitude operations on bucket boom trucks,an improved lightweight armor detection algorithm is proposed based on the PyTorch framework.Firstly,lightweight network GhostNet was used to replace Backbone of YOLO v5s and reduce the complexity of the network and the number of parameters.Secondly,to address the problem of occlusion encountered by workers in high-altitude work,the occlusion situation is simulated by combining Random erase and Mosaic data enhancement.Finally,the loss function is optimized,and the commonly used IOU is replaced by Distance-IOU to solve the problem of large position deviation of the network detection frame.By comparing the results of the improved algorithm and other algorithms on armor detection,the mAP and model size are used to evaluate the experimental results to verify the effectiveness of the improved method and meet the needs of mobile deployment.

关 键 词：轻量化 防具检测 YOLO v5网络 

分 类 号：TB391.41[一般工业技术—材料科学与工程]