基于深度学习的校园安全监控方法研究  

Research on Campus Ecurity Monitoring Method Based on Deep Learning

作　　者：陈文博 孙华志[1] 袁映暄 田佳谊 薛笛晨 CHEN Wenbo;SUN Huazhi;YUAN Yingxuan;TIAN Jiayi;XUE Dichen(Tianjin Normal University,Tianjin 300387,China)

机构地区：[1]天津师范大学,天津300387

出　　处：《信息与电脑》2021年第8期71-76,共6页Information & Computer

基　　金：天津市级大学生创新创业训练计划项目资助(项目编号:202010065045)。

摘　　要：为保护学生人身安全,降低学生因聚集发生事故的概率,笔者设计一种基于卷积神经网络和密度分布特征的校园人群安全监控系统。该系统首先将场景中的人群依据密度进行划分,对于学生稀疏的情况,利用FasterRCNN对目标区域进行判别和分类;针对遮挡密集人群,在人群聚集处使用CSRNet进行拥挤人群密度计算;针对口罩识别,采用Yolov3进行口罩佩戴识别。该系统结合多种算法和方法将人数监测和口罩佩戴功能相结合,及时反馈预警,以提升校园安全管理水平。In order to protect the personal safety of students and reduce the probability of accidents due to clustering of students,the author designed a campus crowd safety monitoring system based on convolutional neural network and density distribution characteristics.The system first divides the crowds in the scene according to density.For sparse students,Faster RCNN is used to identify and classify the target area;for crowds with dense occlusions,CSRNet is used to calculate the crowd density in crowded places;for mask recognition,Use Yolov3 for mask wearing recognition.The system combines multiple algorithms and methods to monitor the number of people and the function of wearing masks,and provide timely feedback and early warning to improve the level of campus safety management.

关 键 词：深度学习 校园安全 监控 人群密度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]