基于DNN压缩感知的无线分布式野外实时监测系统  

A distributed wireless field real-time monitoring based on DNN compressed sensing

作　　者：宋宜轩 宫新保[1] SONG Yi-xuan;GONG Xin-bao(Department of Electronic Engineering,Shanghai Jiaotong University,Shanghai 200240,China)

机构地区：[1]上海交通大学电子工程系,上海200240

出　　处：《信息技术》2022年第8期88-92,共5页Information Technology

摘　　要：基于DNN的压缩感知技术可以有效提升图像压缩率和恢复图像的质量,但是较高的计算复杂度限制了该技术在实际场景中的应用。针对野外实时监测的应用场景,文中提出了一种分布式无线架构,有效克服了有线监测系统布设难度高、灵活性差的问题;同时设计了一种基于FPGA流处理的DNN实现方式,缓解了系统性能功耗比低、实时性差的问题。实验结果证明,针对实时的视频流,该系统的处理时延不到9ms,性能功耗比是GPU的6.17倍。Compressed sensing technology based on DNN can effectively improve the image compression rate and the quality of the restored image.However,the high computational complexity limits the application of this technology in the actual scene.Therefore,a distributed wireless architecture is proposed in this paper for real-time monitoring in the field,which effectively overcomes the problems of high difficulty and poor flexibility of wired monitoring system.A DNN implementation method based on FPGA stream processing is also proposed to alleviate the problems of high-power consumption and high processing delay.The experiment results show that for real-time monitoring video,the processing delay of the system is less than 9ms,and the performance power ratio is 6.17 times better than GPU.

关 键 词：压缩感知 深度神经网络 FPGA 流处理 

分 类 号：TN919.8[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]