基于压缩感知的农情监测节点稀疏采样决策方法  

Sparse Sampling Decision-making Based on Compressive Sensing for Agricultural Monitoring Nodes

作　　者：赵刚 饶元[1] 朱军[1] 李绍稳[1] Zhao Gang;Rao Yuan;Zhu Jun;Li Shaowen(College of Information and Computer Sciences,Anhui Agricultural University,Hefei 230036,Anhui)

机构地区：[1]安徽农业大学信息与计算机学院,安徽合肥230036

出　　处：《长江大学学报（自然科学版）》2019年第1期79-87,I0005,I0006,共11页Journal of Yangtze University(Natural Science Edition)

基　　金：国家自然科学基金项目(61572260);农业部引进国际先进科学技术948项目(2015-Z44;2016-X34);安徽省自然科学基金项目(1608085QF126)

摘　　要：将压缩感知应用于农情监测节点的稀疏采样,能够有效减少数据冗余和能耗,而如何甄选测量矩阵、稀疏基和重构算法,是实现节点稀疏采样和数据收集质量控制的关键。系统分析了基于压缩感知的单节点稀疏采样与重构方法,设计了基于压缩感知的农情监测节点稀疏采样决策系统功能架构,阐述了Python语言环境下系统的实现技术与效果。系统初步应用表明,对于30d连续监测、采样间隔为15min的花卉植株茎流、土壤湿度数据的压缩感知,宜选择固定采样率,测量矩阵、稀疏基、重构算法分别选取周期测量矩阵、差分矩阵和SL0算法,每2～10d重构一次数据可实现数据恢复代价与精度的良好折衷。The application of compressed sensing in sparse sampling of agricultural monitoring nodes could effectively reduce data redundancy and energy consumption.How to select measurement matrix,sparse basis and reconstruction algorithm was the key to realize sparse sampling and data collection quality control.The method of sparse sampling and reconstruction of single node based on compressed sensing was analyzed systematically.The functional framework of sparse sampling decision system based on compressed sensing was designed.The implementation technology and effect of the system in Python language environment were described.The preliminary application of the system shows that for 30 days of continuous monitoring,sampling interval of 15 minutes of flower stem flow,soil moisture data compression perception,should choose a fixed sampling rate,measurement matrix,sparse base,The reconstruction algorithm selects the periodic measurement matrix,the difference matrix and the SL0 algorithm,and reconstructs the data every 2-10 days to achieve a good compromise between the cost and accuracy of data recovery.

关 键 词：压缩感知 农情监测节点 稀疏采样 决策 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] S24[自动化与计算机技术—计算机科学与技术]