孙夏恩,颜王亮






摘 要: 根据疫情期间各场所的防疫信息化、智能化需求,设计了一种基于视频分析与深度学习的疫情防控系统。从口罩检测和人群计数角度,在多场景下通过实时聚合监控和风险预警,配合现场语音反馈和后台数据记录,提升疫情防控管理效能。系统运用了RTMP、HTTP、WebSocket三种应用层协议,结合缓冲区、双进程、多线程设计,利用深度学习网络完成了对最新缓冲帧的实时检测,最终基于Flask框架实现客户端交互。
关键词: 疫情防控; 深度学习; 人群计数; 口罩检测; 视频分析; RTMP
中图分类号:TP391.4;TP183;R184.1 文献标识码:A 文章编号:1006-8228(2022)06-54-04
Design of epidemic prevention and control system based on
video analysis and deep learning
Sun Xiaen, Yan Wangliang
(College of Computer Science and Technology, Hangzhou Dianzi University, Hangzhou, Zhejiang 310018, China)
Abstract: According to the information and intelligent requirements of epidemic prevention in various places during the epidemic period, an epidemic prevention and control system based on video analysis and deep learning is designed. From the perspective of mask detection and crowd counting, real-time aggregate monitoring and risk warning are realized in multiple scenarios. With on-site voice feedback and background data recording, the management efficiency of epidemic prevention and control has been improved. Three application-layer protocols including RTMP, HTTP and WebSocket are used. Combined with designs of buffer, concurrent process and multithreading, the real-time detection of the latest buffer frame is completed by using the deep learning network. Client interaction is implemented based on the Flask framework.
Key words: epidemic prevention and control; deep learning; crowd counting; mask detection; video analysis; RTMP
0 引言
科學技术是人类同疾病较量的锐利武器,人类战胜大灾大疫离不开科学发展和技术创新[1]。疫情大数据地图、出入校人脸识别系统、健康码等疫情背景下诞生的应用,都很好地体现了科学技术在现代社会疫病预防和控制中发挥的重要作用。与此同时,深化计算机视觉、生物特征识别、监控图像目标提取等技术的应用也是推进社会治理模式创新与智慧城市建设的重要举措[2],在疫情防控常态化背景下,智能化与信息化的方法具有低风险、低人力消耗、结果精确等众多优势,有效维护了疫情期间的社会秩序。
本文构建一种基于视频分析与深度学习的疫情防控系统,依托多场景监控资源,利用深度学习技术,部署人群计数、口罩检测模型,并配合直播视频传输技术,构建高效的缓冲策略和异步检测流程,实现了多场景的聚合监控和检测能力,为管理人员和普通流动人员提供双端反馈,充分发挥监控视频的价值,提升场所疫情防控能力。……