【作者】 李建辉；

【导师】 廖桂平；

【作者基本信息】 湖南农业大学 ， 作物信息科学， 2014， 博士

【摘要】 精准、实时、稳定和便捷地获取包括温度、光照度、相对湿度、土壤成分含量、C02浓度、风速和大田图像信息等多种作物生长环境信息因子,对于作物生长调控,乃至现代农业的可持续发展起着越来越重要的作用。无线传感器网络(WSN, Wireless Sensor Networks)技术是实现作物生长环境信息现场感知的一种有效方法,但在近年的应用中暴露出诸多技术问题：有限且不可再生的WSN能量供给机制与长时间、持续的现场监测需求之间形成了突出的矛盾；随着感知对象种类的丰富和发展,现场感知产生的数据量迅速膨胀,对基于轻量级传感器的现场感知业务的稳定运行构成了严峻挑战;由于缺乏适合作物大田环境下的无线信道参考模型,阻碍了人们对基于WSN的现场感知业务的精准分析和进一步优化。另外,针对基层农业单位迫切需求的使用方便、集成多种感知能力的低成本一体化传感器设备的研制,仍然处于起步阶段。针对上述技术问题,本文以油菜及其生长大田环境作为研究样本,从理论层面对延长现场感知业务的生存周期、提升现场感知数据处理能力、构建作物生长大田环境下的信道模型展开了深入研究,并对低成本、一体化的无线感知设备进行了原型产品设计和试制等应用技术探索,旨在为基于WSN的作物生长环境信息现场感知方法提供低成本、低功耗和高效率的关键技术支持,为建立科学的作物生长调控机制奠定基础。1、针对基于WSN的作物生长环境信息现场感知方法的持续、长效运行需求,提出了基于MIMO的最大生存周期方法。研究中以新兴的虚拟MIMO为技术基础,结合MAC子层上的拓扑控制新方法,提出了一种WSN最大生存周期的路由策略(MLRV),获得了数字调制阶数(b=4)、分簇规模(m=9)和虚拟天线阵列规模(nt=3)等影响WSN能量效率的最优参数。通过对比实验,在节点存活量、有效包交付数量和WSN生存周期三个指标上,MLRV具有明显优势,实证了MLRV在延长基于WSN的作物生长环境信息现场感知业务生存周期上的有效性。2、针对基于WSN的作物生长环境信息现场感知方法的稳定性问题,提出了基于马尔可夫链概率模型进行节点状态调度的小波变换数据融合方法。针对轻量级无线传感器的实际运行特征,本文提出了一种以马尔可夫链概率模型进行节点状态调度的小波变换数据融合策略(DCWM)。DCWM利用节点状态变化过程的马尔可夫链模型建立节点状态调度管理机制,以维持较低的路由中断概率,并提出利用单层分解和压缩采样理论,降低压缩编码的开销,提高压缩效率。通过公开数据集的仿真测试和大田对比试验,结果表明,以温度、光照度、相对湿度数据和数字图像为感知样本时,DCWM在压缩精度(2%～4.1%)、压缩率(当测量矩阵规模m=400时,压缩率约20%)、压缩比(采用haar小波时压缩比为3.71,使用wbarb. mat压缩比为4.85)、吞吐量与系统总能耗指标上均优于对比的样本,为现场感知业务的稳定运行提供了核心关键技术支持。3、针对因特定农业环境下信道模型缺乏而造成的对现场感知业务的优化与分析难题,建立了油菜大田环境下WSN信道传播损耗特征模型。以油菜及其生长大田为例,依据不同的生育期提炼出三个典型的测试场景,采用两种载波频率的无线传感器节点测试大田现场的信号损耗值,对测定结果进行了最小二乘法拟合建模,结果表明,供试大田环境下的WSN信道传播损耗特性与传统的移动信道模型存在较大差异,适合采用多斜率对数阴影衰落经验模型进行表征。以节点的有效传播距离和路径损耗预测值为指标,与Free-space、Cost321和Egli等经典模型比较,对拟合模型进行有效性验证,结果表明,针对有效传播距离的预测,单斜率拟合模型的预测偏差绝对值小于4.8m,双斜率模型不超过4.5m；针对路径损耗值的预测,单斜率拟合模型的平均误差小于2.1dB,双斜率模型预测低于2dB,并且780M节点的误差更小,平均误差为1.7dB,优于2.4G的1.9dB。本模型为优化和设计较高能量效率的现场感知业务系统奠定了良好的基础。4、针对部署微型自动气象站和商用ZigBee产品成本高、开放性较差和集成度低等问题,设计并实现了一种基于WSN的低成本、多种感知参数嵌入的油菜生长环境数据采集系统。本课题试制了传感器硬件原型产品并编写了配套的上位机组态软件,通过仿真实验和大田试验,结果表明,温度采集有效精度达到±0.5℃,相对湿度的测量精度达到±3%RH,光强采集范围为1～655351x,满足农业现场环境数据的较高测量要求。在标称电源供电情况下,系统实际有效生存周期超过227天,由于采用AT89C51和nRF2401作为基础硬件平台,成本低廉,有助于大规模部署和应用。更多还原

【Abstract】 To precisely, real-time, stably and portably obtain all kinds of environmental factor information on crop growth, such as temperature, illumination, relative humidity, soil composition and content, CO2density, wind speed and graphic information of big cropland, plays increasingly important role in crow growth control and even sustainable development of modern agriculture.WSN, Wireless Sensor Networks, is an effective method to realize on-site sensing for environment information on crop growth. But many technical problems arise from recent application. Outstanding contradictions come into being between limited non-renewable WSN energy supply system and long-time continuous on-site monitoring requirement. Data volume from on-site sensing has been rapidly increased with richness and development of sensing subject types, which severely challenges stable operation of on-site sensing business, based on light-level sensors. With lack of wireless channel reference model suitable for big cropland environment, accurate analysis and further optimization based on WSN on-site sensing business will be held up. And it is still in early stage for study and development of easy-to-use and low-cost sensor integration equipment, which is integrated with many sensing methods and urgently needed by basic-level agricultural units. This thesis uses oilseed rape and its growth big cropland as study sample, to further study theoretically on extending life cycle of on-site sensing business, improving data processing ability for on-site sensing, building channel model under big field of crop growth, to explore applicative technology of prototype design and development for low-cost and wireless sensing integration equipment, which provides with low-cost, low-power-consumption and high-efficiency key technical support for WSN-based on-site sensing methods of crop growth environmental information, and constitutes a solid basis to build scientific crop growth control mechanism.How to effectively extend life cycle of WSN-based on-site sensing methods of crop growth environmental information, is core problem to keep on-site sensing business on-going and long-lasting.Collecting activity of environmental factor information on crop growth, is usually a long-term, continuous monitoring process with huge demand for energy consumption. But sensor nods planted in the fields is always sealed, small-sized, energy limited and non-renewable as common features. Moreover, intensive crop stem leaf will bring serious signal path loss and frequency selective fading, speed up nod power to run out earlier, and shorten life cycle of on-site sensing business. To blindly increase nod density, improve antenna gain, or use external energy facilities of solar energy etc, will increase deployment cost, speed up network energy consumption and affect normal field work.To effectively extend WSN life cycle is equal to extend life cycle of on-site sensing business. With exsiting problems, based on new virtual MIMO technology and combined with topology control new method on MAC sub-layer, this thesis brings up a WSN maximum life cycle routing policy, to obtain optimal parameters of digital modulator (b=4), cluster scale (m=9) and virtual antenna array scale (nt=3), which affect WSN energy efficiency.By comparative experiments, MLRV has obvious advantage in three indexes of nod survival, effective package delivery quantity and WSN life cycle, which proves that MLRV extends on-site sensing life cycle of crop growth environmental information effectively.Proper data fusion methods are used to compress increasing data scale, which will keep it running stable for WSN-based on-site sensing business.As sensing data of crop growth environmental information has been increasing, light-level wireless sensor nods cannot dynamically and real-time process these data whether in energy resource or calculation resource. This contradiction will result in nod shutdown and offline and then sensing process will be interrupted. In another way, huge redundant data will waste precious energy and shorten life cycle of sensing process.So this thesis brings up wavelet transform data fusion policy, which dispatches nod status with Markov chain probability model. Using Markov chain model of nod status changing, DCWM builds nod status management system, to keep lower routing interruption probability. And it brings up single layer decomposition and compress sampling theory to lower compression encoding expense and improve compression efficiency, to solve light-level sensor calculating storage and energy expense problems.By simulation test of public data sets and big field comparative experiments, the results is that, with temperture, illumination, relative humidity data and digital graphic as sensing sample, DCWM is better than comparison sample in compression accuracy (2～4.1%), compressibility (when measurement array scale m=400, it is about20%), compression ratio (It is3.71while using haar wavelet. It is4.85while using wbarb.mat), handling capacity and system total energy consumtion index. And it provides with core key technology support for stable operation of on-site sensing business. Modeling wireless channel transmission loss characteristics under crop growth big field environment, is further optimization of on-site sensing method of WSN-based crop growth environmental information, and precondition for further design.Taken oilseed rape and its growth big field as example, according to different growth season, three typical test scenes are chosen. Signal loss value has been tested by wireless sensor nod using two kinds of carrier frequency. Test result has been put in a fit modelling with least square method. It shows that there is a big difference bwtween WSN channel transmission loss characteristics under testing big field environment and traditional mobile channel model, so it is suitable to characterize it with empirical model of multiple slope logarithmic shadow fading. Taking effective transmission distance of nods (dma) and predictable value of path loss as index, comparing to traditional models of Free-space, Cost321, Egli and so on, verifying effectively with fit model, we have found that prediction accuracy of fit model has obvious advantage.With the prediction of dmax within characterization range suitable for single-slope and double-slope fit model, absolute value of predicted deviations of single-slope fit model is less than4.8m, and that of double-slope model is not more than4.5m. With the prediction of path loss value, mean error of single-slope fit model is within2.1dB and that of double-slope is less than2dB. And predicted deviations of780M nod is even less, mean error is1.7dB, better than1.9dB of2.4G.This is a good basic for the testing model to optimize and design on-site sensing business of higher energy efficiceny.With the problems of high cost and bad openness for micro auto weather station and commercial ZigBee products, a low-cost WSN-based data collection system of oilseed rape growth environment has been designed and realized. This thesis trial-produces sensor hardware prototype product and writes supporting upper computer state software. By simulation experiment and big field test, the result is that temperature collecting accuracy comes up to±0.5℃at most, humidity testing accuracy is±3%RH, light intensity collecting range is1～65535lx, concurrent data transfer comes up to36channels, which can satisfy higher measurement requirements of agricultural on-site environmental data. With nominal power supply, actual effective systematic life cycle exceeds227days. Low cost will help large-scale deployment and application, thanks to AT89C51and nRF2401as basic hardware platform.更多还原