精准农业宽幅网络高通量组织与非劣分层优化关键技术研究

In precision agriculture, omnidirectional sensor nodes for environmental monitoring coexist with directional sensor nodes for crop disease sensing, which makes the network has the wide-sensing and high-throughput characteristics. It is extremly difficult to balance all the performances such as network connection, blind spot tracking, regional enhancement, multiscale detection, and power management. In allusion to the problem mentioned above, considering multi-hop communication omnidirectional network and directional coverage of image, a boundary effect mode of the hidden blind spots and a multiobjective non-dominated sorting algorithm will be proposed to improve environmental monitoring accuracy of heterogeneous network and to avoid coverage loss. By analyzing characteristics of dynamic network in agricultural environment, such as information density, link quality, delay constraint, a feedback topology based control policy of combinational cognitive network will be presented to realize flexibly adjust network capacity, energy consumption and reliability under different requirement standard. To break through the bottle neck of performance such like bandwidth and anti-interference brought in by multimedia informational sensing, a random field maximal clique concurrent opportunistic cooperative transmission mechanism will be stated and realize high-throughput transmission of the network. In order to solve the problem of network resource competition with directed image and omnidirectional environment data, a high and low frequency perceptive observation matrix for multimedia information is established. Data sparse mapping algorithms and energy equalization mechanism are studied to achieve network load reduction and prolong the network life cycle. In conclusion, this project will provide the theory base including wide-sensing network coverage and high-throughput transmission in precision agricultural monitoring.

精准农业环境监测全向感知节点与作物生理病害有向感知节点并存，具有典型宽幅感知与高通量组织特性，综合平衡网络连通、盲区跟踪、区域增强、多尺度检测、能量控制极为困难。针对以上问题，充分考虑多跳通信全向网络和图像有向覆盖条件，研究作物密集遮挡隐性盲区边界效应模型和多目标非劣分层方法，提高异构网络环境监测精度并避免覆盖缺失；分析生境动态网络信息密度、链路质量、时延约束等特性，研究混合认知网络拓扑反馈控制策略，实现不同要求标准的网络容量、功耗和可靠性灵活调整；为突破多媒体信息感知引入带来网络带宽、抗扰等性能瓶颈，研究随机场极大团并发机会协同传输机制，实现网络高通量传输；为解决不同信息密度的有向图像和全向环境数据网络资源竞争难题，建立多媒体信息高低频感知观测矩阵，研究信号同态稀疏降维与节点能量消涨平抑机制，实现网络减载与网络生命周期延长。项目研究为精准农业生产监测提供宽幅网络覆盖与高通量传输理论基础。

针对精准农业环境监测全向感知节点与作物生理病害有向感知节点并存特点，项目重点研究典型宽幅感知与高通量组织网络连通、盲区跟踪、区域增强、多尺度检测和能量控制等关键技术，创立了一套面向农业网络高通量组织与性能优化理论方法，取得了作物密集遮挡有向隐性盲区侦测与非劣分层部署、农业复杂环境概率链路拓扑认知反馈调度、农业宽幅网络多并发机会协同传输机制、渐变农田高通量网络传输减载与能量配给标度策略等创新成果，主要包括：①创新了作物密集生长条件下的有向覆盖度量函数与区域最小覆盖模型，设计了感知方向、感知视角与作物位置、遮挡视角间自适的覆盖增强部署方法，采用双层禁忌分层搜索方法，确定满足最小覆盖约束的非劣解集，实现遮挡条件下的区域连续覆盖。②提出了农业复杂环境不稳定链路下的无线网络拓扑认知调度方法，采用改进的超模博弈模型对接入频谱与功率进行联合优化控制，实现农田环境自适的认知无线网络拓扑调度，提高了能效利用率与传输可靠性。③设计了信道机会连通与移动机会连通双重条件下的多并发协同传输机制，建立节点密度、网络剩余能量、路径动态损耗、移动节点遍历路径等多参数综合的传输代价模型，采用张量分解方法进行分包关系与机会路径组合选取，提高路径组合选择效率与数据传输通量。④创新了多模态互补表示学习方法，利用离散小波变换、稀疏压缩等建构数据多视角投影降维方法，构建节点能量消涨模型，设计了可再生能源下均衡供电策略。项目成果有效推动了精准农业环境与作物表型生理高效感知及降耗传输前端理论发展，对于无线传感器网络技术在现代化农业生产管控中的应用具有重要指导意义。.依托项目资助发表学术论文37篇，其中SCI检索21篇（一区11篇、二区6篇）、EI检索9篇；申请国家发明专利7项（已授权3项）；获得软著2项，参编专著1部；项目负责人吴华瑞研究员入选国家级人才计划，获得全国农牧渔业丰收奖一等奖1项、河北省科学技术进步奖二等奖1项、吴文俊人工智能科技进步奖三等奖1项。项目成果入选2020年全国农业十大引领性技术、2022年中国农业农村重大科技新成果。圆满完成了项目既定研究任务和考核指标。