基于字典学习和物联网边缘计算的水利水电监测图像压缩传输研究  

The Study on Compression Transmission of Water Conservancy and Hydropower Monitoring Image Based on Dictionary Learning and Edge Computing of Internet of Things

作　　者：郭翔 王永涛[2] 余云昊 狄查美玲 GUO Xiang;WANG Yong-tao;YU Yun-hao;DI Zhai-meiing(Power Dispatching Control Center of Guizhou Power Grid Co.,Ltd.,Guizhou Guiyang 550002,China;Guizhou Provincial Water Conservancy Research Institute,Guizhou Guiyang 550002,China)

机构地区：[1]贵州电网有限责任公司电力调度控制中心,贵州贵阳550002 [2]贵州省水利科学研究院,贵州贵阳550002

出　　处：《机械设计与制造》2023年第10期26-30,35,共6页Machinery Design & Manufacture

基　　金：贵州省水利厅水利技术示范项目([2019]SF-201913)。

摘　　要：针对水利水电工程中图像压缩传输的需求,这里提出基于K-SVD、压缩采样匹配追踪(CoSaMP)重构的欠完备字典学习及稀疏表达方式,并结合图像块字典更新学习得到终端边缘计算有损压缩方法。之后设计了应用于4G、GPRS网络环境的物联网终端系统。工程应用结果表明,所提的方法及设计的终端模块能够实现在短时间(计算时间最长14.2s)内达到最高93%、最低59.8%的数据压缩比,重建图像能够清晰表达原图像特征,峰值信噪比(PSNR)达到(24.3~25.4)db,优于传统压缩传输方法。这里所提系统具有低成本、低功耗、较高性能的优点,有效解决实际应用中图像采集连接超时、成像时延大、数据传输成本高的问题。Countering the demand of image compression for water conservancy and hydropower engineering image transfer,an incomplete dictionary learning and sparse representation based on K-SVD and CoSaMP reconstruction method was proposed and combined the way with block dictionary updating learning obtained the terminal edge computing method.Then,the Internet of things terminal system applied in 4G,GPRS network environment is designed.The engineering application results showed that the proposed method and the terminal module could achieve the highest 93%and the lowest 59.8%data compression ratio in a short time(the longest calculation time was 14.2s),the reconstructed image could clearly express the characteristics of the original im-age,and the Peak Signal Noise Ratio(PSNR)reaches(24.3~25.4)db,which is better than the traditional method.The proposed system had the advantages of low cost,low power consumption and high performance,which could effectively solve the problems of net connection timeout,large imaging delay and high data transmission cost in practical application.

关 键 词：图像压缩 K-SVD CoSaMP 字典学习 稀疏表达 物联网边缘计算 

分 类 号：TH16[机械工程—机械制造及自动化] TP39[自动化与计算机技术—计算机应用技术] TN92[自动化与计算机技术—计算机科学与技术] TV86[电子电信—通信与信息系统]