基于改进YOLOv7的安全帽佩戴检测算法
Algorithm of Safety Helmet Wearing Detection Based on Improved YOLOv7
杨大为 1张成超1
作者信息
- 1. 沈阳理工大学 信息科学与工程学院,沈阳 110159
- 折叠
摘要
为提高工作场所安全帽佩戴的检测精度,提出一种基于YOLOv7 网络架构的改进算法.首先,在特征提取网络中引入卷积块注意力机制(CBAM)取代YOLOv7 中主干网络部分原有的卷积模块(CBS),增强网络的特征提取能力,加强网络对目标和背景的分辨能力;其次,为解决由于网络层数的加深导致小目标特征减弱甚至消失的问题,增加一个小目标层,通过将浅层网络特征与深层网络特征融合,进一步保留小目标特征.实验结果表明,原YOLOv7 对安全帽佩戴检测的均值平均精度为86.1%,改进后到达93.4%,实现了检测精度的提高.
Abstract
In order to improve the detection accuracy of safety helmet wearing in workplace,an improved algorithm based on YOLOv7 network architecture is proposed.Firstly,an convolutional block attention module(CBAM)is introduced into the feature extraction network to replace the o-riginal convolutional module(CBS)in the backbone network of YOLOv7,so as to enhance the feature extraction capability of the network and enhance the recognition ability of the target and background.Secondly,in order to solve the problem that the small target features weaken or even disappear due to the deepening of the number of network layers,a small target layer is added to further preserve the small target features through the integration of shallow network features and deep network features.The mean average precision of the original YOLOv7 detection model for safety helmet wearing is 86.1%.The experimental results show that detection accuracy has been im-proved to 93.4%.
关键词
安全帽/特征提取网络/注意力机制/小目标Key words
safety helmet/feature extraction network/attention mechanism/small target引用本文复制引用
基金项目
辽宁省教育厅高等学校基本科研项目(LJKMZ20220612)
出版年
2024
NETL
NSTL
万方数据