基于改进YOLOv5模型的复杂场景口罩佩戴识别

Mask Wearing Recognition in Complex Scenes Based on Improved YOLOv5 Model

为满足疫情时期的特殊需要,基于改进YOLOv5模型,设计一款应用于复杂场景的口罩佩戴识别检测系统。系统采用Py Charm集成开发环境,从网络上爬取1600张口罩佩戴相关的图片,在原始k-means算法基础上加入聚类算法,获取与真实框之间的更高的先验框。在人机交互界面使用Qt组件设计,图像和模型数据加载使用开源OpenCV视觉库实现。口罩佩戴检测的核心算法使用目标检测算法中的YOLOv5训练并结合PyTorch框架实现。实验结果表明,系统能够实现多人场景下人群佩戴口罩状况的检测,在多人物以及实时检测场景中准确率有所提高,在使用和效果上都具有自身特有的优势。

In order to meet the special needs during the epidemic period, a mask wearing identification and detection system for complex scenes is designed based on the improved YOLOv5 model. The system adopts Py Charm integrated development environment, crawls 1600 pictures related to mask wearing from the network, and adds clustering algorithm to the original k-means to obtain a higher prior frame with the real frame. Qt component is used to design the human-computer interaction interface, and open source OpenCV visual library is used to load images and model data. The core algorithm of mask wearing detection is implemented by YOLOv5 training in the object detection algorithm combined with PyTorch framework.The experimental results show that the system can detect people wearing masks in multi-person scenes,and the accuracy is improved in multi-person and real-time detection scenes, which has its own unique advantages in use and effect.