深度学习在作物病害图像识别方面应用的研究进展

在作物生产管理过程中,正确及时地诊断作物所患病害非常关键.基于深度学习的图像识别为作物病害自动快速诊断提供了新途径.相比传统图像识别所用的模式识别方法,深度学习网络模型能自行提取特征且能够由低维特征抽象出高维特征,取得更好的学习效果.系统梳理了深度学习在图像自动化识别方面的发展历程,介绍了浅层神经网络的相关概念,阐述了...     

天津铜锣湾广场深基坑开挖流固耦合分析

通过对天津铜锣湾广场深基坑开挖过程的流固耦合分析,揭示了主、被动土压力和孔隙水压力的分布规律,探讨了开挖前期降水对土体变形的影响,验证了该深基坑工程设计的合理性。所得结论,对类似工程有重要的理论指导意义和借鉴价值。     

Effect of subsoiling on soil properties and winter wheat grain yield

A 2‐year field experiment was carried out in loessal soil in a semi‐humid climate to explore the integrative influences of subsoiling on soil properties and winter wheat grain yield. Results showed that it was essential to deepen the plough pan in loessal soil. The sharply increased soil penetration resistance (>7 MPa) in plough pan under dry soil condition was especially harmful for plant growth. In rotary tillage treatment, the waterlogging conditions caused by shallow plough pan slightly produced macropores and preferential water flow. Significantly, higher grain yield was obtained in the subsoiling tillage treatment, increased by 21.9% in 2016 and 11.3% in 2017, respectively. Subsoiling tillage improved the resilience of winter wheat under adverse climate conditions. Subsoiling tillage should be popularized in loessal soil with shallow plough pan in semi‐humid climate. This study may provide valuable information on soil sustainable use and management in loessal soil.     

Efficient Purification of R-phycoerythrin from Marine Algae (Porphyra yezoensis) Based on a Deep Eutectic Solvents Aqueous Two-Phase System

R-phycoerythrin (R-PE), a marine bioactive protein, is abundant in Porphyra yezoensis with high protein content. In this study, R-PE was purified using a deep eutectic solvents aqueous two-phase system (DES-ATPS), combined with ammonium sulphate precipitation, and characterized by certain techniques. Firstly, choline chloride-urea (ChCl-U) was selected as the suitable DES to form ATPS for R-PE extraction. Then, single-factor experiments were conducted: the purity (A₅₆₅/A₂₈₀) of R-PE was 3.825, and the yield was 69.99% (w/w) under optimal conditions (adding 0.040 mg R-PE to ChCl-U (0.35 g)/K₂HPO₄ (0.8 g/mL, 0.5 mL) and extracting for 20 min). The sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) results revealed that the purified R-PE contained three main bands. One band was presented after purification in native-PAGE. The UV-vis spectra showed characteristic absorption peaks at 495, 540, and 565 nm. R-PE displayed an emission wavelength at 570 nm when excited at 495 nm. All spectra results illustrated that the structure of R-PE remained unchanged throughout the process, proving the effectiveness of this method. Transmission electron microscope (TEM) showed that aggregation and surrounding phenomena were the driving forces for R-PE extraction. This study could provide a green and simple purification method of R-PE in drug development.     

基于计算机视觉的养殖动物计数方法研究综述

数量计量是动物养殖管理的基础工作,其结果对于动物养殖的生产效率、养殖成本管控及经济效益评估等具有重要意义。基于计算机视觉的计数方法解决了传统人工计数存在的测量误差大、耗时费力等问题,减轻了养殖人员的工作负担。本文统计分析了近十年的养殖动物视觉计数相关研究,从传统机器学习与深度学习两方面对养殖动物计数算法进行分析与讨论。此外,对水产养殖、畜禽养殖与特种动物养殖领域的养殖动物计数应用进行梳理与总结。同时,对目前公开发布的养殖动物计数数据集进行概述。最后,从数据集、应用场景、计数方法3方面分析讨论养殖动物计数研究面临的主要挑战,并对未来研究进行展望。     

基于LSTM模型预测不同水保工程措施条件下土壤侵蚀量——以辽西北地区为例

[目的] 准确模拟和预测不同水保工程措施下土壤侵蚀量,为辽西北地区精准预测土壤侵蚀量提供技术和理论依据。[方法] 基于野外径流小区2011—2021年的监测数据,包括：最大30 min和60 min降雨强度(I₃₀和I₆₀)、降雨历时(T)、降雨量(P)和土壤侵蚀量,建立了长短期记忆神经网络(LSTM)分别对3种工程措施(水平槽、果树台田和梯田)下的土壤侵蚀量进行预测。并将LSTM预测结果与3个经典机器学习模型〔反向传播神经网络(BP)、随机森林(RF)和支持向量机(SVM)〕预测的结果进行对比。[结果] ①在3种工程措施中,I₃₀,I₆₀,T和P对土壤侵蚀量的影响程度不同,但I₃₀,I₆₀和T对土壤侵蚀量的影响大于P。②利用BP模型预测土壤侵蚀量的相对均方根误差(NRMSE)均大于0.2。③相比于RF和SVM模型,LSTM模型在3种工程措施下(水平槽、果树台和梯田)预测土壤侵蚀量的NRMSE分别降低了约0.04～0.08,0.02～0.08,0.05～0.08。④利用I₃₀和T作为LSTM模型的输入特征预测土壤侵蚀量的精度与使用I₃₀,I₆₀,T和P为输入特征时的预测精度相近。[结论] 在辽西北地区3种水保工程措施中,利用LSTM模型基于最大30 min雨强和降雨历时对土壤侵蚀量进行预测,取得了较其他传统模型高的预测精度。这说明LSTM模型可在同类地区土壤侵蚀量的精准模拟和确定水土保持措施中推广和应用。     

基于红外热成像和改进YOLO v5的作物病害早期识别

为实现作物病害早期识别,本文提出一种基于红外热成像和改进YOLO v5的作物病害早期检测模型,以CSPD-arknet为主干特征提取网络,YOLO v5 stride-2卷积替换为SPD-Conv模块,分别为主干网络中的5个stride-2卷积层和Neck中的2个stride-2卷积层,可以提高其准确性,同时保持相同级别的参数大小,并向下阶段输出3个不同尺度的特征层;为增强建模通道之间的相互依赖性,自适应地重新校准通道特征响应,引入SE机制提升特征提取能力;为减少模型计算量,提高模型速度,引入SPPF。经测试,改进后YOLO v5网络检测性能最佳,mAP为95.7%,相比YOLO v3、YOLO v4、SSD和YOLO v5网络分别提高4.7、8.8、19.0、3.5个百分点。改进后模型相比改进前对不同温度梯度下的作物病害检测也有提高,5个梯度mAP分别为91.0%、91.6%、90.4%、92.6%和94.0%,分别高于改进前3.6、1.5、7.2、0.6、0.9个百分点。改进YOLO v5网络内存占用量为13.755MB,低于改进前基础模型3.687MB。结果表明,改进YOLO v5可以准确快速地实现病害早期检测。     

深水油气管道再启动过程模拟计算

为了研究深水油气管道再启动过程的多相流动和传热规律,结合传热学和多相流理论,建立了深水管道再启动过程数学模型,给出了模型的定解条件和求解方法;利用该模型,仿真模拟了深水管道再启动过程中温度和压力等参数的变化规律;在此基础上对水合物的生成情况进行预测并分析参数敏感性。研究表明：启动流量、绝热层的导热系数和厚度可以明显改变管内流体的温度分布,且启动流量越大、绝热层导热系数越小、厚度越大、管内流体温度越高,生成水合物的风险越小。研究结论可为深水油气管道再启动的安全进行提供理论指导。     

Combining vision information and tag information to extract Deep Web result pages content

Extracting content from deep web pages is a challenging problem due to the underlying intricate structures of such pages. A vision and tags based approach (DVS) is proposed. It primarily utilizes the vision information and tag information on the Deep Web result pages to extract the content structure of pages. This approach includes two steps as follows: First, the vision information and tag information are produced by analyzing the Cascading Style Sheet and the DOM Tree to generate an initial visual tree of the Deep Web result page. And then, the Path Shingle (PS) algorithm is employed, by considering both of the vision and the tag information, and the blocks in the visual tree are clustered according to the similarity computing result of them to produce the final visual tree, i.e., the content structure of pages. The innovations of DVS are that it utilizes the vision information and tag information on the Deep Web pages to extract the content structure; and stores the vision information as a tree to tansform the analysis of the vision information to a vision attribute tree. Experiments are conducted with a large set of Web databases called UIUC’s TEL. The experimental results show that the vision and tag based approach has high precision compared with the WTS algorithm and the VIPS algorithm.