基于Jetson,Nano和改进YOLOv7算法的安全帽佩戴目标检测技术分析
Analysis of Helmet Wearing Object Detection Technology Based on Jetson,Nano,and Improved YOLOv7 Algorithm
姚冲 1邓在辉1
作者信息
- 1. 武汉纺织大学计算机与人工智能学院,湖北 430200
- 折叠
摘要
阐述YOLOv7-tiny算法在安全帽佩戴目标检测数据集上检测精度较低,同时在光照条件差、背景模糊、目标受遮挡等情况下容易出现错检、漏检等问题.为此,针对性地加入CBAM注意力机制对模型进行改进,将特征图先输入通道注意力模块,提升模型对目标物体的关注度,再使用空间注意力机制使模型能够更好地学习到目标的位置信息.通过对YOLOv7-tiny算法的网络结构改进,取得了91％精度,对比YOLOv7-tiny(90.5％)精度提升了0.5％.经过模拟测试,发现改进后的模型能够有效减少错检、漏检概率,并且能够在低算力平台上实现高精度,低误判的部署,从而有效实现安全帽佩戴监管.
Abstract
This paper describes that the YOLOv7 tiny algorithm has low detection accuracy on the target detection dataset for helmet wearing,and is prone to false positives and missed detections under poor lighting conditions,blurred background,and obstructed targets.To this end,a targeted CBAM attention mechanism is added to improve the model.The feature maps are first input into the channel attention module to enhance the model's attention to the target object,and then the spatial attention mechanism is used to enable the model to better learn the position information of the target.By improving the network structure of the YOLOv7 tiny algorithm,a 91％accuracy was achieved,which was 0.5％higher than the YOLOv7 tiny(90.5％)accuracy.After simulation testing,it was found that the improved model can effectively reduce the probability of false positives and missed detections,and can achieve high-precision,low misjudgment deployment on low computing power platforms,thereby effectively achieving supervision of helmet wearing.
关键词
目标检测/YOLOv7/Jetson/Nano/注意力机制/安全帽Key words
object detection/YOLOv7/Jetson Nano/attention mechanism/safety helmet引用本文复制引用
出版年
2024
NETL
NSTL
万方数据