基于多行人复杂环境下的社交距离检测研究  

Social Distance Detection Based on Monocular Vision and Pedestrian Detection in the Multi-Object Complex Conditions

作　　者：吴瑰[1] 曾鹏[1] 陶俊[1] Wu Gui;Zeng Peng;Tao Jun(Jianghan University,Wuhan 430056,China)

机构地区：[1]江汉大学,湖北武汉430056

出　　处：《湖北成人教育学院学报》2023年第2期73-76,81,共5页Journal of Hubei Adult Education Institute

基　　金：武汉市教育局教育研究计划项目“VR技术在新工科实践教学模式中的实施与应用”(2019068);江汉大学武汉研究院开放基金“十四五期间武汉市人工智能驱动新经济发展的调查研究”(IWHS20212017)。

摘　　要：保持安全的社交距离在预防高传播风险疾病时起着至关重要的作用,为规范行人间的社交距离,减小行人间可能存在的新冠传播风险,提出一种基于单目视觉的多行人测距方法,实现对监控视角下的多行人间距测量。该方法通过行人检测模型捕捉监控视角下的多行人目标,利用单目测距原理,实现了多行人间的距离度量,通过分析这些距离来达到规范行人间的社交距离。实验结果表明,该方法能够高效和迅速检测出不符合社交距离规范的人群。Maintaining a safe social distance plays a vital role in preventing high-risk diseases.In response to the current outbreak of COVID-19,it is necessary to regulate the social distance between pedestrians and reduce the possible risk of COVID-19 transmission between pedestrians.Then,it is helpful to propose a single-based system.The visual multi-pedestrian distance measurement method could realize the measurement of the multi-pedestrian distance under the surveillance angle.This method uses the pedestrian detection model to capture the multi-pedestrian target under the monitoring perspective.Similarly,it uses the monocular ranging principle to realize the distance measurement between the multi-pedestrians.The aim is to achieve the standardized social distance between pedestrians by analyzing these distances.The experimental results show that the method could detect people who do not satisfy the social distance norms efficiently and quickly.

关 键 词：社交距离 单目测距 相机标定 行人检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]