摘要
本文针对传统的YOLOv5模型提出有效的结构优化,减小了设备体积,增加了预测速度,适用于算力低、内存少的嵌入式平台,并在边缘计算设备Jetson Nano上成功部署,提升了检测速度。基于Jetson Nano和YOLOv5算法,本文提出了一种新型的火灾自动报警系统。该系统能够通过视频监控实时感知火灾现场,对火源进行有效识别,及时发出报警信号,有效减少火灾事故的发生。实验结果表明,该系统具有高效、准确的火灾检测和识别能力,能够满足实际应用需求。此外,本文所提方案不仅可以对YOLOv5模型进行高效的压缩与加速,其他卷积神经网络模型都可参照本文方案进行操作模块的替换,为其他深度学习算法部署在资源有限的嵌入式平台上提供借鉴。
In this paper,we propose an effective structural optimization for the traditional YOLOv5 model to reduce the device size and increase the prediction speed,which is suitable for embedded platforms with low computing power and low memory,and successfully deployed on the edge computing device Jetson Nano to improve the detection speed.Based on Jetson Nano and YOLOv5 algorithm,a novel automatic fire alarm system is proposed in this paper.The system is able to sense the fire scene in real time through video monitoring and effectively identify the fire source,send out alarm signals in time,and effectively reduce the occurrence of fire accidents.The experimental results show that the system has efficient and accurate fire detection and identification capabilities,and can meet the practical application requirements.In addition,the scheme proposed in this paper can not only efficiently compress and accelerate the YOLOv5 model,but also other convolutional neural network models can refer to the scheme in this paper for the replacement of the operation module,which provides a methodological reference for other deep learning algorithms deployed on embedded platforms with limited resources.
作者
尹昊
Yin Hao(Ningxia Architectural Design Institute Co.,Ltd,Yinchuan 750002,China)
出处
《智能建筑电气技术》
2024年第3期98-105,共8页
Electrical Technology of Intelligent Buildings
