基于PID神经网络的滑翔增程制导炮弹解耦控制系统设计

采用Wiener空间的两个算子以及相关的恒等式，提出了新的方法证明了关于高斯过程函数的中心极限定理，并给出了该中心极限定理的应用实例.     

Performance of DSSAT-Nwheat across a wide range of current and future growing conditions

Crop models are widely used in agricultural impact studies. However, many studies have reported large uncertainties from single-model-based simulation analyses, suggesting the need for multi-model simulation capabilities. In this study, the APSIM-Nwheat model was integrated into the Decision Support System for Agro-technology (DSSAT), which already includes two wheat models, to create multi-model simulation capabilities for wheat cropping systems analysis. The new model in DSSAT (DSSAT-Nwheat) was evaluated using more than 1000 observations from field experiments of 65 treatments, which included a wide range of nitrogen fertilizer applications, water supply (irrigation and rainout shelter), planting dates, elevated atmospheric CO₂ concentrations, temperature variations, cultivars, and soil types in diverse climatic regions that represented the main wheat growing areas of the world.DSSAT-Nwheat reproduced the observed grain yields well with an overall root mean square deviation (RMSD) of 0.89 t/ha (13%). Nitrogen applications, water supply, and planting dates had large effects on observed biomass and grain yields, and the model reproduced these crop responses well. Crop total biomass and nitrogen uptake were reproduced well despite relatively poor simulations of observed leaf area measurements during the growing season. The low sensitivity of biomass simulations to poor simulations of leaf area index (LAI) were due to little changes in intercepted solar radiation at LAI >3 and water and nitrogen stress often limiting photosynthesis and growth rather than light interception at low LAI.The responses of DSSAT-Nwheat to temperature variations and elevated atmospheric CO₂ concentrations were close to observed responses. When compared with the two other DSSAT-wheat models (CERES and CROPSIM), these responses were similar, except for the responses to hot environments, due to different approaches in modeling heat stress effects.The comprehensive evaluation of the DSSAT-Nwheat model with field measurements, including a comparison with two other DSSAT-wheat models, created a multi-model simulation platform that allows the quantification of model uncertainties in wheat impact assessments.     

A Receiving Surface of Perturbative Photon Fluxes in Interaction Between Gravitational Waves and EM Fields

According to Electrodynamics in curved spacetime, the interaction between gravitational waves (GW) and electromagnetic (EM) fields will produce perturbative electromagnetic fields. In a special EM system, namely, a Gaussian beam (Gb) propagating through a static magnetic field, the photon flux along the direction of the electric field of the Gb will be only a pure perturbative photon flux,when a GW propagates along the symmetric axis of a Gaussian beam. Detailed alanysis and calculation of the photon flux along several typical surfaces are performed, and a possible optimum plane which can receive perturbative photons up to 10~4s~(-1) is proposed. It is possible to provide an important method to detection of GW.     

The River Wadu role play — ten years experience

This paper describes a role playing planning exercise that has been used for training agriculturists, engineers and agricultural economists for more than ten years. It sets out the objectives and key features of the exercise and the experience that has been gained in its use. The exercise has been found to successfully test and integrate subject matter expertise acquired from formal teaching. It also highlights the multiple and often conflicting objectives of irrigation planning, and the need for full evaluation of alternatives and sound technical judgements under conditions of inadequate data and time pressure. The exercise is also successful in developing professional personal skills such as group working, interviewing and presentation, and is recommended as an effective training aid.     

基于高斯过程建模的物联网数据不确定性度量与预测

物联网已经成为农业大数据最重要的数据源之一,自动观测数据的质量控制对农业生产分析以及基础科研数据应用非常重要。针对农业物联网观测的一类非平稳时间序列数据中的数据缺失、野值剔除、感知故障预警和长时间预测等问题,采用光滑弱假设高斯先验,构建了基于高斯过程的自回归模型表征的动态系统,并通过样本集学习,形成能考虑噪声干扰的传感变化规律建模,并可提供预测误差带用于预测数据的不确定性度量。针对原始数据的缺失和野值问题,采用基于高斯过程的短期预测,可补齐缺失数据,利用其不确定性度量可甄别数据野值,进行野值剔除与替换,并在此基础上判断感知故障;给出了基于输入数据不确定性传播的多步迭代预测方法,使长期预测仍可以跟踪农业数据的动态轨迹,并可为其预测值提供不确定性度量;将温室采集的真实传感数据用于分析试验,验证了高斯过程用于服务器端的农业时间序列数据采集质量控制的可行性。     

基于对数相似度约束Otsu的自然场景病害果实图像分割

针对自然场景下,由于复杂背景以及多变环境,水果病害果实图像分割难的问题,提出了一种基于对数相似度约束Otsu和水平集活动轮廓的近椭圆形病害果实图像分割方法。考虑背景的复杂多变,提出对数相似度约束Otsu分割来区分病害果实与背景;由于水平集活动轮廓模型的局部最优性,提出采用自适应膨胀系数的改进距离规则水平集活动轮廓模型来精确演化轮廓。先对病害果实区域样本的颜色进行混合高斯建模,获得整个病害果实图像与样本模型的对数相似度;对对数相似度进行约束Otsu阈值分割以及形态学滤波;采用最小二乘法对滤波后的曲线轮廓进行椭圆拟合,对拟合后的椭圆采用自适应膨胀系数的距离规则水平集活动轮廓演化,得到病害果实完整轮廓。对18个不同场景的病害果实进行分割,平均误判率和漏判率分别为1.77%和1.6%,实验结果表明,该方法可以从复杂自然场景图像中分割出病害果实。     

基于高光谱的双季稻分蘖数监测模型

旨在阐明双季稻分蘖数与冠层反射高光谱间的定量关系,构建基于高光谱的双季稻分蘖数监测模型。基于不同早、晚稻品种和施氮水平的田间试验,于关键生育期(分蘖期、拔节期和孕穗期)测定早、晚稻分蘖数,同步使用FieldSpec HandHeld 2型高光谱仪采集早、晚稻冠层反射高光谱数据,分别利用光谱指数法和连续小波变换构建新型光谱指数和敏感小波特征对双季稻分蘖数进行监测,建立双季稻分蘖数光谱监测模型,并用独立试验数据进行检验。结果表明,新型光谱指数和敏感小波特征对双季稻分蘖数的监测效果优于其他类型光谱参数(植被指数和“三边”参数),其中位于红边区域的小波特征db7(s9,w735)监测早稻分蘖数时表现最优,监测模型R2为0.754,模型检验相对均方根误差RRMSE为0.128;位于红边区域的小波特征mexh(s6,w714)监测晚稻分蘖数时表现最优,监测模型R2为0.837,模型检验RRMSE为0.112。研究结果可为双季稻分蘖数快速无损监测和群体质量精确调控提供理论基础与技术支持。     

基于高斯函数的春玉米叶片功能期模型构建与应用

叶片功能期是影响光合生产能力的关键因素,冠层叶片功能期的量化评估对玉米植株生长和产量形成具有重要意义。本研究于2015—2017年在中国农业科学院作物科学研究所吉林公主岭试验站进行,定株观测先玉335和郑单958两个品种各个叶位叶片展开时间和衰老时间,基于2015年和2016年试验数据,以高斯函数(y =a+b×e-(x-c)^2/2d^2)模拟玉米各叶位叶片功能期的动态变化,并用2017年数据验证,在此基础上进一步明确了模型特征参数的生理学意义,简化了叶片功能期模型构建的方法。研究条件下利用高斯函数构建的玉米叶片功能期模型年际间稳定性好、品种间区分度大。进一步解析利用一阶导(功能期最大值)、二阶导(功能期变化速率最大的点)、三阶导(功能期开始快速增大的点)等于零取整后的叶位并配合最顶部叶位(n)和基部第1叶这5个转折叶位叶片功能期构建的模型拟合度良好,极大地简化了该模型参数拟合的数据需求,并探讨了利用该模型函数对玉米叶片功能分组的可能性。本研究为玉米生产能力的量化分析提供了思路和方法,对各类玉米生长模型的完善和其他相关研究也有借鉴意义。     

基于Lab颜色空间的非监督GMM水稻无人机图像分割

为解决传统水稻冠层图像分割算法性能在很大程度上依赖于训练数据集的质量,且分割效果易受田间多变光照强度影响,导致水稻生产参数估计精度不高等问题,提出一种基于Lab颜色空间的非监督贝叶斯方法,用于田间水稻无人机图像分割.模型参数从每个独立、未标记的无人机图像直接学习获得,无需训练.不同图像会有不同的模型参数,该算法能够适应...     

基于机器视觉的奶牛发情行为自动识别方法

及时检测奶牛发情、适时人工授精、减少空怀奶牛,是奶牛养殖场增加产奶量的关键手段。针对基于运动量和体温等体征的接触式奶牛发情识别方法会造成奶牛应激反应且识别准确率不高的问题,提出了一种非接触式奶牛发情行为自动识别方法。该方法首先使用改进的高斯混合模型实现运动奶牛目标检测,然后基于颜色和纹理信息去除干扰背景,再利用AlexNet深度学习网络训练奶牛行为分类网络模型,识别奶牛爬跨行为,最终实现对奶牛发情行为的自动识别。在供试数据集上的试验结果表明,本文方法对奶牛发情的识别准确率为100%,召回率为88.24%。本文方法可应用于奶牛养殖场的日常发情监测中,为生产管理提供辅助决策。