摘要
设计了全向移动式智能安防机器人。该机器人采用轻量化改进后的YOLOv5检测算法,并应用ArcFace人脸识别损失函数,基于FreeRTOS系统建立报警任务和巡逻监控任务,具有巡逻监控、身份识别、自主报警等功能。为满足机器人目标识别的实时性,通过快速下采样方法重新对YOLOv5检测算法的Backbone层和Neck层进行设计,并分别对输入图片进行卷积和最大池化操作,并将二者的结果相融合传入下一阶段,使网络的运算速度比之前提升了2.322倍。对机器人进行了安防效果测试,结果表明,机器人具有较高的识别精准度和较强的主动预防能力。
To improve the warning ability,target recognition,and tracking of traditional surveillance and security systems,a fully mobile intelligent security robot was designed.The robot adopts a lightweight,improved YOLOv5 detection algorithm combined with the ArcFace face recognition loss function scheme,and establishes alarm and patrol monitoring tasks based on the FreeRTOS system.It has functions such as patrol monitoring,identity recognition,and autonomous alarm.In order to meet the real-time target recognition requirements,the Backbone layer and Neck layer of the YOLOv5 detection algorithm were redesigned using a fast down-sampling method.The input images were convolved and max-pooled separately,and the results were fused and passed to the next stage,resulting in a 2.322-fold increase in network operation speed.The security effect of the robot was tested,and the results show that it has high recognition accuracy and strong proactive prevention capabilities.
作者
马晨
苏易衡
张紫钰
杨昌儒
马玉凯
李和洋
MA Chen;SU Yiheng;ZHANG Ziyu;YANG Changru;MA Yukai;LI Heyang(School of Instrumentation Science and Opto-Electronics Engineering,Beijing Information Science and Technology University,Beijing 100192,China)
出处
《实验室研究与探索》
CAS
北大核心
2023年第7期94-98,103,共6页
Research and Exploration In Laboratory
关键词
安防机器人
全向移动
巡逻监控
轻量化
security robot
omnidirectional movement
patrol monitoring
lightweight