基于语义分割的非结构化田间道路场景识别

环境信息感知是智能农业装备系统自主导航作业的关键技术之一。农业田间道路复杂多变,快速准确地识别可通行区域,辨析障碍物类别,可为农业装备系统高效安全地进行路径规划和决策控制提供依据。该研究以非结构化农业田间道路场景为研究对象,根据环境对象动、静态属性进行类别划分,提出一种基于通道注意力结合多尺度特征融合的轻量化语义分割模型。首先采用Mobilenet V2轻量卷积神经网络提取图像特征,将混合扩张卷积融入特征提取网络最后2个阶段,在保证特征图分辨率的基础上增加感受野并保持信息的连续性与完整性;然后引入通道注意力模块对特征提取网络各阶段特征通道依据重要程度重新标定;最后通过空间金字塔池化模块将多尺度池化特征进行融合,获取更加有效的全局场景上下文信息,增强对复杂道路场景识别的准确性。语义分割试验表明,不同道路环境下本文模型可以对场景对象进行有效识别解析,像素准确率和平均像素准确率分别为94.85%、90.38%,具有准确率高、鲁棒性强的特点。基于相同测试集将本文模型与FCN-8S、SegNet、DeeplabV3+、BiseNet模型进行对比试验,本文模型的平均区域重合度为85.51%,检测速度达到8.19帧/s,参数数量为,相比于其他模型具有准确性高、推理速度快、参数量小等优点,能够较好地实现精度与速度的均衡。研究成果可为智能农业装备在非结构化道路环境下安全可靠运行提供技术参考。     

设施菜地土壤氮素运移及淋溶损失模拟评价

设施菜地因大水大肥管理方式导致的氮素淋失已成为当前关注焦点。探寻氮素淋失阻控技术需要首先探明土壤中NO_3~--N的运移和淋失过程,找到淋失阻控的关键点,从而实现蔬菜栽培高产量低环境成本。本研究以京郊设施菜地黄瓜-番茄轮作系统为研究对象,通过田间试验获取土壤温度、湿度、NO_3~--N含量等数据,对反硝化-分解(DNDC)模型进行参数校验,并以农民常规种植模式为基线情景,设置改变土壤基础性质、灌溉量、施氮量等不同情景,运用DNDC模型对设施菜地系统土壤氮素运移及淋溶损失进行定量评价。结果表明:经验证后的DNDC模型能够较好地模拟蔬菜产量、5 cm土壤温度和0～20 cm土壤孔隙含水率变化以及NO_3~--N的迁移过程,是模拟和评价氮素运移和损失的有效工具。模拟不同情景表明,设施菜地0～60 cm土壤NO_3~--N累积主要受灌溉水量和氮肥施入量的影响,此外土壤pH和土壤有机碳的变化也是影响NO_3~--N运移的重要因子。节水节肥是设施菜地氮素淋失减量的最有效方法,相比常规措施,同时减少20%灌溉量和20%施氮量可明显降低59.04%的NO_3~--N淋失量。同时,在节水节肥的基础上改变灌溉方式并提高20%土壤有机碳含量,在保证蔬菜产量的前提下,能够进一步降低69.04%的NO_3~--N淋失量。可见, DNDC模型为设施菜地NO_3~--N淋失评价和阻控提供了一个较好的解决方案。在当前重点关注减氮节水等管理措施的同时,提高土壤本身的质量,不失为一种更有效的减少设施菜地氮素淋失的途径。     

Emission of dimethyl sulfide,carbonyl sulfide,and carbon bisulfide from paddy fields

Abstract

Emission rates of dimethyl sulficle (DMS), carbonyl sulfide (COS), and carbon disulfide (CS₂) to the atmosphere from paddy fields at Ryugasaki, Ibaraki Prefecture, Japan were measured by using the closed chamber method. DMS was emitted mainly through the rice plant and its emission rate was much higher than those of COS and CS₂. During a cropping period COS was slightly absorbed by the rice plant. Significant diurnal and seasonal variations of DMS iuxes were observed. The highest DMS iux was observed immediately after the heading day of rice plant. The annual DMS emission rate was in the order of mineral plot > straw plot > no-N plot, ranging from 4.5 to 6.9 mg S m^-2 yr^-1. The annual COS emission rate was in the order of straw plot > no-N plot > mineral plot, ranging from -0.2 to 1.8 mg S m^-2 yr^-1. The annual CS₂ emission rate was slightly higher in the straw plot, ranging from 0.9 to 2.0 mg S m^-2 yr^-1.     

Influence of regional-scale soil heterogeneity and land use on water discharge formation in the “Stettiner Haff” catchment in North-East Germany

Abstract

The influence of regional soil distribution and land use on discharge formation in the “Stettiner Haff” catchment (8,456 km²) was investigated. The model ABIMO was used to calculate mean discharge for a long period (1961 – 1990) in consideration of regional differences in hydrological site conditions and soil heterogeneity. The highest discharge was found on settlement areas (276 mm a^?1), on devastated plots (319 mm a^?1) and on the non-hydromorphic agriculturally used sandy soils (214 mm a^?1). Due to their small size these sub-areas were found to influence regional discharge for a limited extent only. On loam and clay soils only about 100 mm a^?1 of discharge was formed. However, this share proved to be most important for the height of regional discharge because of the high area proportion in the region. Mineral soils of comparable parent material produce lower discharge under hydromorphic conditions than under non-hydromorphic conditions. With increasing clay content of the soils these differences decreased. The agriculturally used areas were found to be most important for the region's water discharge. They accounted for 83% of the discharge, i.e. they originated 119 mm a^?1 on average. At forest sites, the discharge height was about 52 mm a^?1. Sites close to groundwater table (groundwater level <1 m) proved to act as depletive areas.     

Effect of long-term application of fertilizers on soil quality and rice yield in a salt-affected coastal region of India

The studied soils were collected from a 24-year long-term fertilizer experiment with rice (Oryza sativa L.) grown on tropical Typic Endoaquepts during monsoon season only. The organic carbon (OC) and total nitrogen (TN) status of the soils improved with the N₁₀₀P₂₂K₄₂ treatment than the initial values at the start of the experiment. The available P status of the soils depleted in the N₁₀₀P₁₁K₂₂ and N₀P₀K₀ treatments than the initial value. Soil microbial biomass carbon (MBC) and basal soil respiration (BSR), activities of urease, acid and alkaline phosphatases as well as fluorescein diacetate hydrolyzing activity (FDHA) were the highest in the N₁₀₀P₂₂K₄₂ treatment. The highest percentage of MBC in OC in the N₁₀₀P₂₂K₄₂ treatment revealed the accumulation of organic matter with an overall improvement in nutrient status of soil, since MBC is a labile pool of plant nutrients. Long-term application of fertilizers (N₁₀₀P₂₂K₄₂) had no deleterious effect on microbial and biochemical soil quality parameters. Rice grain yields were inconsistent with N₁₀₀P₀K₀ and N₀P₀K₀ while that of N₁₀₀P₂₂K₄₂, N₁₀₀P₁₁K₂₂ and N₁₀₀P₂₂K₀ increased gradually during the course of the experiment for each block period of six years. The grain yield of rice showed high significant correlation with the microbial and biochemical soil parameters.     

中国低产田状况及改良策略

从新阶段我国农业、特别是土壤学发展的需求出发,总结了国内外有关低产田方面的研究成果。从"低产田"的界定入手,对我国低产田的状况及其负面影响、低产田形成的自然要素和人类活动的影响以及主要障碍因子等进行了系统分析,并初步测算了低产田改良的粮食增产潜力等。我们认为"低产田"可以定义为"在现有的正常耕作栽培管理技术水平条件下,因为耕地本身存在的障碍或者限制因子,导致作物生长发育差,产量较当地高产田低30%以上,且年际间变异大的农田"。我国低产田具有面积大、类型多,而且分布相对集中、障碍因子较明显等特征,低产田主要的障碍因子包括:有机质贫乏、养分匮乏或失衡、土壤酸化、土壤盐渍(碱)化、沙化、土壤板结、潜育化、表土大量流失、严重干旱、多因子组合。在此基础上,最后从我国现代农业发展的迫切需求出发,提出了今后一段时间内低产田改良研究的重点及有关政策和对策建议。     

旱地全膜覆土穴播和全沙覆盖平作对小麦田土壤水分和产量的调节机理

以春小麦品种‘陇春27’为试材,采用田间试验法,以裸地平作为对照,研究半干旱区旱地全膜覆土穴播和全沙覆盖平作对小麦田土壤水分和产量的调节作用。结果表明:与裸地平作(CK)相比,全膜覆土穴播(PM)和全沙覆盖平作(SM)小麦田0~40 cm土壤水分条件明显改善,尤其在干旱年份,能满足小麦前期生长,同时促进小麦出苗后对0~200 cm土壤水分的利用;种植第1年PM在60~80 cm土层耗水量最大,SM和CK在40~60 cm土层耗水量最大;种植第2年PM以120~180 cm土层耗水量最多,SM和CK则以60~80 cm土层耗水量最多。连续种植两年后,PM耗水深度从120 cm延伸到200 cm,SM耗水深度从120 cm延伸到140 cm,CK耗水深度无变化;小麦田休闲效率PM最大,SM次之,CK最小,但是各处理休闲效率随种植年限增加而降低。可见,PM和SM能改善小麦前期生长水分环境,促进出苗后耗水,并加快小麦对土壤深层水分的利用,因而与CK相比,PM产量增加48.77%~815.79%,SM产量增加49.41%~702.24%。但随种植年限增加,耗水深度加大,休闲效率降低,多年种植可能对土壤水分生态产生不利影响。     

基于DTW距离的拖拉机传动轴载荷样本长度计算方法

针对工程领域载荷样本长度确定方法对农业机械载荷适用性较差的问题,提出了一种基于动态时间扭曲(DTW,Dynamic Time Warping)距离计算载荷样本长度的方法。首先,运用无线扭矩传感器获取犁耕作业下拖拉机传动轴的动态载荷,结合传动轴犁耕作业下的载荷特点,划分作业工况;再将作业循环叠加并经过归一化处理后进行DTW距离计算,确定传动轴载荷样本长度;最后,计算参数外推中均值混合正态分布函数的拟合参数相对误差。结果表明,运用DTW距离方法确定的载荷样本长度,能够使均值双正态分布函数拟合得到的参数相对误差保证在10%以内。近似均值精度估计法和均值曲线拟合法的拟合参数相对误差最大值分别为126.06%和80.62%,而基于DTW距离计算方法的拟合参数相对误差最大值为5.43%,验证了基于DTW距离的载荷样本长度计算方法的适用性。研究结果可为拖拉机传动轴田间试验载荷样本长度确定提供参考。     

气体传感器鉴别花椒产地研究

目前花椒产地鉴别基本以感官评定为主,缺乏客观性,在实施应用时难以做到量化和标准化,难以做出判断。因此设计研发一种快速鉴别花椒的智能装置。该装置以气体传感器阵列为核心,能够独立对花椒气味信息进行检测和鉴别,区分不同产地的同类花椒。利用主成分分析和WilksΛ统计分析对检测数据进行处理。提取主成分5个,累积贡献率为94.41%,其对应Fisher判别模型训练集平均准确率达到88.6%,验证集90%,WilksΛ统计分析最终选取8个变量,其对应判别Fisher模型训练集平均准确率91.82%,验证集95%。对WilksΛ统计所选取变量建立细分类交叉验证的Fisher判别模型,平均正确率达到97.27%,将模型移植到采集装置,完成智能花椒品种鉴别装置。该方法是一种简便高效的花椒品种鉴别方法,可为今后进一步研究花椒产地、分级提供检测仪器和理论依据。     

竖壁贴附送风改善冷藏库内流场特性

果蔬贮藏期间,冷藏库内的气流分布是影响果蔬贮藏品质和送风能量利用率的关键因素,合理的气流组织对冷藏库内流场分布是至关重要的。该研究以某苹果冷藏库为研究对象,采用竖壁贴附送风方式,并对其流场建立了三维稳态的SST湍流模型来研究库内流场的分布特性以及货物的冷却效果,同时与传统的冷风机直吹送风的流场特性进行了对比。结果表明：在相同送风量下,相比冷风机直吹送风方式,竖壁贴附送风使得库内温度不均匀系数和速度不均匀系数分别降低了31%和47%。而且竖壁贴附送风使库内空气与货物的换热更加充分,进而使得送风的能量利用率提高了19%。因此,竖壁贴附送风方式应用于冷藏库是一种有效且节能的气流组织方式。