基于YOLOv5s+DeepSORT的公共场所行人体温监测方法

Pedestrian Temperature Monitoring in Public Places Based on YOLOv5s+DeepSORT

在人员密集、人流量大的公共场所,发热人员可能携带传染性病毒。公共场所通常使用非接触式体温检测摄像头检测行人体温。然而现有设备在疫情防控中存在无法追踪发热人员导致防疫效率低的问题。对此,提出了一种将体温检测设备、YOLOv5s框架与改进的DeepSORT算法相结合以监测行人体温的方法。该方法首先采用体温检测摄像头获取行人的脸部检测框与体温信息,并采用YOLOv5s对画面中的行人进行精准识别;然后使用所提出的IOU-PR匹配算法对关联信息进行转换,获取行人检测框与体温的关联信息;最后结合改进的DeepSORT算法实现对画面中行人的实时追踪。结果表明,所提出的方法能够稳定地检测行人体温并保持追踪,提高防疫工作效率。

This paper proposes a method that integrates temperature detection devices,the YOLOv5s framework,and an improved DeepSORT algorithm to monitor pedestrian body temperature.Initially,this approach utilizes the thermal camera to capture the facial detection box and temperature information of pedestrians,accurately identifying them in the image using YOLOv5s.Subsequently,the IOU-PR matching is utilized to transfer associated information,obtaining the correlation between pedestrian detection boxes and temperature.Ultimately,the improved DeepSORT algorithm enables real-time tracking of individuals with fever in the image.The results demonstrate that the proposed method can stably detect pedestrian body temperature and maintain tracking,thereby enhancing the efficiency of epidemic prevention efforts.