基于深度学习的人群密度检测系统在嵌入式平台上的应用

Application of Crowd Flow Density Detection System Based on Deep Learning in Embedded Platform

人群密度计数是智能化视频监控分析领域的关键要素和研究热点。围绕基于目标检测和边缘计算的人群密度检测系统进行研究与设计,主要分为硬件端、通信层和Web端3个部分。硬件端利用摄像头实时监测,树莓派作为边缘设备运行YOLOv4-tiny的目标检测算法,将结果送入通信层-数据库,Web端利用AJAX等异步技术获取数据,引入PyMySQL库完成Web后端与数据库的连接,利用Apache Echarts框架进行JavaScript的交互功能设计,将系统在实际场景测试。结果表明,系统识别效果较好,可视化界面实时更新系统参数,实现了在嵌入式平台上的稳定运行。

Crowd density counting is a key problem and research hotspot in the field of intelligent video surveillance and analysis.This paper focuses on the research and design of crowd density detection system based on object detection and edge computing,which is mainly divided into three parts:hardware,communication layer and Web.Hardware uses camera to monitor in real time,raspberry pie as edge device runs YOLOv4-tiny target detection algorithm,and sends the results to communication layer-database.Web uses asynchronous technology such as AJAX to obtain data,introduces PyMySQL library to complete the connection between Web back-end and database,and uses Apache Echarts framework to design JavaScript interactive function.Finally,the system is tested in the actual scene.Experimental results show that the recognition effect of the system is good,and the visual interface updates the system parameters in real time,thus realizing the stable operation on the embedded platform.