采用双模态联合表征学习方法识别作物病害

基于深度卷积神经网络的视觉识别方法在病害诊断中表现出色,逐渐成为了研究热点。但是,基于深度卷积神经网络建立的视觉识别模型通常只利用了图像模态的数据,导致模型的识别准确率和鲁棒性,都依赖训练数据集的规模和标注的质量。构建开放环境下大规模的病害数据集并完成高质量的标注,通常需要付出巨大的经济和技术代价,限制了基于深度卷积神经网络的视觉识别方法在实际应用中的推广。该研究提出了一种基于图像与文本双模态联合表征学习的开放环境下作物病害识别模型(bimodalNet)。该模型在病害图像模态的基础上,进行了病害文本模态信息的嵌入,利用两种模态病害信息间的相关性和互补性,实现了病害特征的联合表征学习。最终bimodalNet在较小的数据集上取得了优于单纯的图像模态模型和文本模态模型的效果,最优模型组合在测试集的准确率、精确率、灵敏度、特异性和F1值分别为99.47%、98.51%、98.61%、99.68%和98.51%。该研究证明了利用病害图像和病害文本的双模态表征学习是解决开放环境下作物病害识别的有效方法。     

基于迁移学习的GoogLeNet果园病虫害识别与分级

为提高果树病虫害危害程度分级精度进而更好地指导果园病虫害防治，采用迁移学习技术与GoogLeNet模型相结合的方法，对6种果园作物的25类病虫害样本进行识别与危害程度分级研究；同时，探究不同数据集大小以及不同优化算法对模型性能的影响；基于MATLAB平台设计了一款可视化的病虫害识别与分级系统。结果表明:1)基于迁移学习的GoogLeNet模型，对病虫害识别精度可达99.35%，危害程度分级精度可达92.78%；2)在相同训练参数下，本研究模型比AlexNet、VGG-16、ResNet-18、SqueezeNet、原GoogLeNet及MobileNet-v2模型验证精度提高了2.38%~11.44%，并且收敛速度最快；3)本研究模型识别精度随着数据集的增大而提高；在3种优化算法中SGDM算法耗时最短且精度最高，更适合本研究模型。通过拍摄果树叶片病害区域图像，本研究设计的系统能够在0.43 s左右准确识别出果树种类、病害类型以及危害等级等信息。     

露地盆栽月季主要病虫害及绿色防控技术

露地盆栽月季在栽培过程中极易受黑斑病、白粉病、霜霉病、灰霉病、月季长管蚜、二斑叶螨、西花蓟马、花蓟马、月季叶蜂等病虫害侵染,使月季花期缩短,花朵变小,植株矮小,严重影响其观赏价值。本文着重论述露地盆栽月季主要病虫害及绿色防控技术。     

Progress in relationship between epithelial sodium channel and related respiratory diseases

Epithelial sodium channel (ENaC) is an ion channel widely distributed in various tissues and organs of human. It is composed of 3 homologous subunits and allows the flow of sodium ion across epithelial cells, maintain?ing water-salt balance in the cells. Recent studies show that abnormal expression or dysfunction of ENaC in the respiratory system affects water-salt balance, fluid transportation and cell mobility, and causes abnormal changes of the airway surface liquid level and impaired clearance. ENaC is closely related to the development of respiratory diseases, such as cystic fibrosis, asthma and chronic obstructive pulmonary disease. This article reviews the progress in ENaC structure, function and roles in related respiratory diseases in order to provide a reference for the treatment of the diseases.     

吐鲁番秋延设施黄瓜病虫害绿色防控技术

为确保吐鲁番生态安全、农产品安全生产和重要农产品的有效供给，对秋延晚设施黄瓜病虫害的绿色防控技术进行探索及总结，主要内容包括农业措施（清洁棚室、高温闷棚、品种选择、合理间作、科学灌溉）、物理措施（防虫网、黄板诱杀）、生物措施（合理利用生物药剂）及化学措施（合理用药），实现了减少化学防治2～8次，667 m2节约成本240元左右，经济、生态、社会效益显著，助力农户节本增收和当地特色农业提质增效。     

北方地区林业病虫害防治探究

我国的林业技术在几十年的探索与研究中取得了较大发展,但病虫害对林业的保持和发展危害巨大,北方地区作为林业发展的重要区域,应认识到病虫害的危害,并明确其发生特点,探究加强防治的有效措施,以减少病虫害对北方地区林业发展的影响。     

2个抗病相关基因与高油酸油菜病害关系分析

为快速筛选抗病性强的高油酸油菜品种,本研究以20个高油酸油菜品系为试验材料,调查不同材料的抗病情况,并通过RT-qPCR测定经miRNA测序筛选得到的与抗病相关基因HSP90和ATG3的表达情况,同时分析其与病情指数的关系。结果表明,20个高油酸油菜品系的发病率与病情指数呈线性正相关关系,且差异极显著;HSP90基因从苗期至花期表达量呈逐渐降低的趋势,角果期有先升高后降低和逐渐升高2种表达模式;ATG3基因从苗期至花期表达量也呈逐渐升高的趋势,角果期表达模式与HSP90基因相同。用HSP90基因在花期-叶中表达量≥1.5倍内参时进行抗病材料筛选,有95%的准确率;用ATG3基因在花期-叶中表达量≥1.8倍内参时进行抗病材料筛选,有80%的准确率,此外,也可用HSP90在花期-叶结合ATG3在五~六叶期-叶中表达情况进行预测,有80%的准确率。本研究结果为高油酸油菜抗病材料的早期筛选和油酸油菜育种提供了一定的理论依据。     

基于部首嵌入和注意力机制的病虫害命名实体识别

为了解决农业病虫害命名实体识别过程中存在的内在语义信息缺失、局部上下文特征易被忽略和捕获长距离依赖能力不足等问题，以农业病虫害文本为研究对象，提出一种基于部首嵌入和注意力机制的农业病虫害命名实体识别模型(Chinese agricultural diseases and pests named entity recognition with joint radical embedding and self attention, RS-ADP)。首先，该模型将部首嵌入集成到字符嵌入中作为输入，用以丰富语义信息。其中，针对部首嵌入设计了3种特征提取策略，即卷积神经网络(Convolutional neural network, CNN)、双向长短时记忆网络(Bidirectional long short term memory network, BiLSTM) 和CNN-BiLSTM；其次，采用多层不同窗口尺寸的CNNs层提取不同尺度的局部上下文信息；然后，在BiLSTM提取全局序列特征的基础上，采用自注意力机制进一步增强模型提取更长距离依赖的能力；最后，采用条件随机场(Conditional random field, CRF)联合识别实体边界和划分实体类别。在包含11个类别和24715条标注样本的农业病虫害自制语料上进行了实验。结果表明，本文模型RS-ADP在该数据集上精确率、召回率和F1值分别为94.16%、94.47%和94.32%；在具体实体类别上，RS-ADP在作物、病害、虫害等易识别实体上F1值高达95.81%、97.76%和97.23%。同时，RS-ADP在草害、病原等难以识别实体上F1值仍保持86%以上。实验结果表明，本文所提模型能够有效识别农业病虫害命名实体，其识别精度优于其他模型，且具有一定的泛化性。     

Effect of relative humidity at either acute or chronic moderate temperature on growth performance and droppings' corticosterone metabolites of broilers

The present study aimed to investigate the effect of relative humidity(RH) at either acute or chronic moderate ambient temperature(Ta) on growth performance and droppings' corticosterone metabolites of broilers.Two experiments were conducted: effect of RH(35,60 or 85%) on average daily feed intake(ADFI) and droppings' corticosterone metabolites at acute(1 d: 20–26 or 31–20°C,26 or 31°C for 6 h d–1 at 10:00–16:00) moderate Ta(experiment 1) and effect of RH(35,60 or 85%) on growth performance and droppings' corticosterone metabolites at chronic(step-wisely increasing temperature by 3°C every 3 d from 20 to 32°C within 15 d: 20–23–26–29–32°C) moderate Ta(experiment 2).Droppings were collected at the 2,4,6,8,and 22 h after Ta-RH controlled in experiment 1 and at the 2,4,6,and 22 h after Ta controlled to 32°C in experiment 2.The results showed that: 1) In experiment 1,85% RH increased(P0.05) the droppings' corticosterone metabolites at the 2,6,8,and 22 h and 35% RH increased(P0.05) it at the 2 and 22 h compared to the 60% RH.Moreover,85% RH further increased(P0.05) it compared to the 35% RH,however,no difference(P0.05) was found in ADFI among the three RH groups at acute moderate 26°C; 35 and 85% RH increased(P0.05) droppings' corticosterone metabolites at the 2,6,8 and 22 h and decreased(P0.05) ADFI compared to the 60% RH,moreover,85% RH further increased(P0.05) droppings' corticosterone metabolites and further decreased(P0.05) ADFI compared to the 35% RH at acute moderate 31°C; and the average of droppings' corticosterone metabolites in the whole period had a negative correlation(P0.02) with the ADFI.2) In experiment 2,85% RH increased(P0.01) droppings' corticosterone metabolites only at the 2 h and decreased(P0.02) ADFI and average daily gain(ADG) compared to the 60% RH,no difference(P0.05) in droppings' corticosterone metabolites was found between the 35 and 60% RH,however,35% RH decreased(P0.01) ADG compared to the 60% RH,and the average of droppings' corticosterone metabolites in the whole period also had a negative correlation(P0.02) with ADFI and ADG.In conclusion,droppings' corticosterone metabolites could be used as a RH stress index and low and high RH,especially high RH,reduced growth performance possibly through inducing RH stress at moderate temperature.