基于改进SSD安全头盔反光衣检测算法  

To Improve the SSD Safety Helmet Reflective Clothing Detection Method

作　　者：韩泽佳 肖秦琨[1] 张立旗 HAN Ze-jia;XIAO Qin-kun;ZHANG Li-qi(School of Armament Science and Technology,Xi'an Technological University,Xi'an 710016,China)

机构地区：[1]西安工业大学兵器科学与技术学院,西安710016

出　　处：《自动化与仪表》2021年第9期63-68,共6页Automation & Instrumentation

摘　　要：施工人员正确穿戴安全头盔和反光衣是进行安全生产和保障人身安全的重要一环,当前对安全头盔反光衣穿戴的检查还是依靠传统人工的方式,存在费时费力问题,针对这种情况,使用了深度学习中的SSD(single shot multibox detector)算法作为基本网络框架实时对目标人物进行无人化穿戴检测,同时针对原SSD算法存在检测精度不高的问题,对原SSD算法进行了改进,首先使用了部分ResNet50网络替换内部的VGG-16作为特征提取网络;其次在SSD算法的高层卷积模块中加入可形变卷积模块,使检测目标时更好地适应目标的不同尺寸来提高检测精度。实验结果表明,该网络结构在检测安全头盔和反光衣上精确度和速度上表现优异。The correct wearing of safety helmet and reflective clothing for construction personnel is an important part of safety production and personal safety.At present,the inspection of safety helmet reflective clothing still relies on the traditional manual way,which is time-consuming and laborious.In view of this situation,SSD(single shot multibox detector)algorithm in deep learning is used as the basic network framework to carry out humanized wearable detection on the target in real time.Meanwhile,the original SSD algorithm is improved to solve the problem of low detection accuracy of the original SSD algorithm.Firstly,part of ResNet50 network is used to replace the internal VGG-16 as the feature extraction network.Secondly,the deformable convolution module is added into the high-level convolution module of SSD algorithm to better adapt to the different size of the target to improve the detection accuracy.The experimental results show that the network structure performs well in the accuracy and speed of detecting safety helmet and reflective clothing.

关 键 词：安全头盔反光衣检测 SSD算法 ResNet50 可形变卷积 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]