摘要
为解决深度学习在实际视频信号分析和应用部署上的不足,提出一种基于边缘端的视频信号分析系统来实现目标检测和行为识别。系统以TPU芯片BM1684为核心,提出了视频分析的实现原理和步骤,介绍了目标检测算法YOLOv5和行为识别算法Co3D CNNs,并将这两种算法部署在该芯片内,同时在实验环境和不同的实际场景中对该系统的性能进行了验证,结果表明本系统能够正确进行多种目标检测和行为识别。
In order to solve the shortcomings of deep learning in actual video signal analysis and application deployment,an edge-based video signal analysis system is proposed to achieve target detection and behavior recognition.The system takes the TPU chip BM1684 as the core,proposes the realization principle and steps of video analysis,introduces the target detection algorithm YOLOv5 and the behavior recognition algorithm Co3D CNNs,and deploys these two algorithms in the chip.The performance of the system is verified in the actual scene,and the results show that the system can correctly perform a variety of target detection and behavior recognition.
作者
张莉
张畅
聂红梅
ZHANG Li;ZHANG Chang;NIE Hong-mei(Library,Hunan Agricultural University,Changsha 410128,China)
出处
《电脑与信息技术》
2022年第5期43-45,共3页
Computer and Information Technology
关键词
深度学习
目标检测
行为识别
边缘端部署
deep Learning
object detection
behavior recognition
edge deployment
