基于CT图像的土壤孔隙结构三维重建及水力学性质预测

为了更好地了解土壤孔隙结构对水分运动过程的影响机制,该文利用黄淮海平原原状潮土CT扫描图像,通过数字图像分析和计算机重建技术对孔隙结构进行三维重建,根据图像分析获得的孔隙大小分布和连通性等形态学参数建立了用于描述孔隙尺度结构特征对水分运动影响机制的网络模型,据此预测了样本尺度(样本体积为385.84cm3)的土壤水力学性质。结果表明,模型预测的水力学性质和实测值基本吻合,变化趋势基本一致,二者的决定系数达0.94以上。结果表明相关网络模型可以较好地模拟孔隙尺度的水分运动过程,可用于预测土壤的非饱和水力学性质。     

基于图像处理的土壤表层含水率在线检测方法研究

土壤表层含水率是影响作物生长的重要因素,在节水灌溉控制系统中对其进行在线检测尤为重要。运用计算机数字图像处理技术,探索一种土壤表层含水率在线检测的新方法。通过土壤表层图像的中值滤波、图像模式转换和“坏区”过滤,提取图像的特征参数——灰度值,并对土壤表层图像的灰度值与其含水率的关系进行了试验研究。理论分析与试验结果表明,土壤含水率的百分比与土壤表层图像的灰度值之间是一种近似线性的函数关系。     

基于多重特征增强与特征融合SSD的荔枝检测

使用无人机拍摄的荔枝图像目标尺寸小、特征信息不足.为了更多、更好地检测到荔枝,该研究提出一种基于多重特征增强与特征融合的SSD(Single Shot Multibox Detector based on Multiple Feature Enhancement and Feature Fusion,MFEFF-SSD...     

Dispersing sandy soil for the measurement of particle size distributions using optical laser diffraction

Despite the speed and accuracy of particle size distribution measurement by optical laser diffraction instruments, the dispersion of the sample, especially of soil aggregates, remains time-consuming. The Malvern Instruments obscuration measurement was used to calculate the dispersion ratio as an indicator of soil disaggregation. It enabled rapid and direct comparison between measurements made using different amounts of soil, dispersed using different fluids and dispersion methods. The dispersion ratio was necessary to select the most suitable media and method of dispersion for the soil samples and was useful for identifying grain breakage and flocculation. A standard for size distribution measurements involved a soil sample suspended in tap water and dispersed using ultrasonic action for 3 min. Reproducibility of grain size distribution measurements on the Malvern laser sizer showed that variation caused by resampling was greater than variation in the optical measurement provided that the measurement sweeps for the instrument were appropriately set. Blending size distributions for material measured using two different focal lenses was problematic. It is suggested that the laser light source Coulter LS-100 may be a more suitable instrument when many sample measurements are required.     

采用改进YoloV4模型检测复杂环境下马铃薯

为解决马铃薯联合收获机在作业过程中分级清选的问题,并在收获作业过程中实时监测评估收获状态,该研究提出一种在光照亮度变化大、土壤与薯块遮挡、机器振动以及尘土干扰等情况下对马玲薯进行识别检测并快速准确获取马铃薯数量以及损伤情况的机器学习模型。在卷积神经残差网络中引入轻量级注意力机制,改进YoloV4检测网络,并将YoloV4结构中的CSP-Darknet53网络替换为MobilenetV3网络,完成特征提取。试验结果表明,基于卷积神经网络的深度学习方法相比于传统Open-CV识别提高了马铃薯识别精度,相比于其他传统机器学习模型,MobilenetV3-YoloV4识别速度更快,马铃薯识别的全类平均准确率达到91.4%,在嵌入式设备上的传输速度为23.01帧/s,模型鲁棒性强,能够在各种环境下完成对正常马铃薯和机械损伤马铃薯的目标检测,可为马铃薯联合收获机智能清选以及智能收获提供技术支撑。     

耕层构造的土壤结构质量-径级数字图像分析

为了探讨以土壤结构体为单元的耕层构造定量方法,该文利用犁耕生成的土壤结构体2D图像计算其质量-径级分布。分别以30°、45°、60°及90°拍摄获取土壤结构体的数字图像,计算土壤结构体各径级区间的几何指标,拟合质量-径级分布模型。结果表明土壤结构体的棱角性和形状指数随径级增大而增加,但矩形度随之减小;以60°拍摄所得的土壤结构体质量-投影面积关系的拟合精度最高,各粒径区间R2均不低于0.89;数字图像筛分与手工测量所得的土壤结构体质量-径级分布无显著差异(P0.05),表明数字图像筛分是从2D投影面信息获取土壤结构体质量-径级分布的准确方法;相对于Weibull和Rosin-Rammler模型,用Gaudin-Schuhmann模型拟合获得的土壤结构体质量-径级分布效果较优,用该模型拟合数字图像筛分所得的土壤结构体质量-径级分布,R2为0.98;相对于干筛法,数字图像筛分方法的划分的径级区间更精细,所得的模型拟合精度更高。     

采用改进YOLOv4-Tiny模型的柑橘木虱识别

黄龙病是一种以柑橘木虱为传播媒介的毁灭性病害,其关键预防措施是在果园现场环境对柑橘木虱识别监测,辅助果农进行早期防治。该研究基于YOLOv4-Tiny模型提出一种适用于嵌入式系统的柑橘木虱识别模型。通过改进YOLOv4-Tiny模型的颈部网络,利用浅层网络的细节信息以提高模型识别柑橘木虱的平均精度;采用交叉小批量归一化（Cross mini-Batch Normalization,CmBN）方法代替批归一化（Batch Normalization,BN）方法,通过累计卷积层的输出,提升统计信息的准确度;针对柑橘木虱易被遮挡的问题,模型训练时使用Mosaic数据增强,提升模型对遮挡目标的识别能力。通过自行建立的柑橘木虱图像数据集完成模型的试验验证。结果表明,该模型的柑橘木虱平均识别精度为96.16%,在图形处理器（Graphics Processing Unit, GPU）上的推理速度为3.63 ms/帧,模型大小为24.5 MB,实现了果园环境下快速准确地识别柑橘木虱,可为黄龙病防治技术的进一步发展提供参考。     

Soil displacement and tillage erosion during secondary tillage operations: the case of rotary harrow and seeding equipment

This study reports the results of a series of experiments that were set up on agricultural land in central Belgium to investigate soil translocation and erosivity resulting from a secondary tillage operation using an implement sequence of a rotary harrow and seeder. Aluminium cubes were used as tracers of soil movement. Results show that soil displacement resulting from tillage with such an implement sequence is far from insignificant. This is mainly related to the relatively shallow tillage depth as well as to the loose initial soil condition of such secondary tillage operations. The calculated value for the tillage transport coefficient k (123 kg m⁻¹ per tillage operation) is comparable with k-values from implements that are considered to be more erosive, like mouldboard and chisel implements. In conclusion, this study shows that tillage erosion not only results from relatively aggressive tillage operations such as mouldboard and chisel passes, but that secondary operations contribute significantly to soil displacement and tillage erosion.     

An evaluation of 2D-image analysis techniques for measuring soil microporosity

To compare soil porosity measurements, manual, semiautomatic and automatic 2D‐image analyses were performed on three sets of images of the same fields. The first and second image sets were obtained, by a fluorescence microscope, on the polished surfaces of soil blocks impregnated with a fluorescent resin and on the thin sections made from them, respectively. A scanning electron microscope in backscattered mode was used to acquire the third set of images on the thin sections. In the manual image analysis procedure, image segmentation was based on the best visual impression and carried out individually for each image with the UTHSCSA Image Tool software. For the semiautomatic method, the software μCTanalySIS was used; the images were segmented by double hysteresis, after interactive selection of the thresholding values. Automatic thresholding, based on an analysis of the image intensity histogram, was performed using the image analysis software Image‐Pro^® Plus. Average microporosity measurements were compared between image types and image analysis methods using marginal regression models for continuous outcomes. The mean area covered by pores was significantly different depending on the type of image and the method of image analysis. Results after the automatic procedure were significantly larger than after manual or semiautomatic thresholding. On the other hand, average porosities were larger for backscattered electron images than for fluorescent images, regardless of the image analysis procedure used. These results stress the need for standardization of image analysis protocols and warn of the dangers of comparing soil porosity measurements performed on different types of images.     

Development and testing of a model for predicting tillage effects on nitrate leaching from cracked clay soils

Both water movement and nitrate leaching in structured soils are strongly influenced by the nature of the macro-porosity. That macro-porosity can however also be manipulated by choice of tillage operations. In order to investigate the potential impacts of tillage on rates of nitrate leaching from structured soils, a model specific to these soils, CRACK-NP was developed. The model, its application and validation for an experimental site on a heavy clay soil (Verti-Eutric Gleysoil) at Brimstone Farm, Oxfordshire, UK, is described. The model considers the soil as a series of aggregates whose size is also the spacing of the macro-porosity. Water and solutes move in the macro-pores, but within the peds they move only by diffusion, internal infiltration and root uptake (evaporation). The model reflects the influence of diffusion limitation in the release of solutes to by-passing water. The model was then used to investigate the influence of variable ped spacings which were created by variations in tillage practices. The results both from the model and from the field data demonstrated that finer soil structures, which have larger surface contact areas and shorter diffusion path lengths, present greater opportunities for interaction between peds and the water moving around them, and so release more nitrates through the drainage waters.     

Precise tillage systems for enhanced non-chemical weed management

Soil and residue manipulation can assist weed management by killing weeds mechanically, interfering in weed lifecycles, facilitating operations and enhancing crop establishment and growth. Current tillage systems often compromise these functions, resulting in heavy reliance on herbicides, particularly in no-till systems. Herbicides are an exhaustible resource, so new approaches to merge soil conservation and non-chemical weed management are needed. This paper broadly reviews various preventive and curative non-chemical weed management tactics. It also demonstrates how innovations can be derived from functional requirements of weed management operations, and from biological processes and weaknesses in weed's lifecycles. Mechanical weeding and enhancement of weed seed mortality are highlighted as examples. Major limitations with mechanical weeding include limited weed control in crop rows at early vulnerable crop stages, weather-dependent effectiveness, and difficulties in handling crop residues. Precise steering and depth control, improved seedbed friability and lighter tractors or controlled traffic could bring considerable improvements. To expose weed seeds to predators, position them for fatal germination, viability loss or low emergence may require completely different soil displacement patterns than those of current implements and systems. Controlled traffic and precise strip tillage offer good opportunities for implementing these weed management strategies in minimum-tillage systems.     

Dynamic programming analysis of pore space

A method based on the principles of dynamic programming is developed and applied to digital soil images to determine tortuosity, effective porosity, and the shortest paths available for flow. Pathways for flow are determined by minimizing a resistance factor. Analysis of two numerically simulated soils finds that a platy soil has greater effective porosity and smaller tortuosity than a layered soil. The ability of the dynamic programming model to analyse large digital images with complex pore structure is subsequently demonstrated by the analysis of a digital image of an agricultural soil. The results from the agricultural soil point to the need for digital soil image analysis to be three–dimensional if convective and diffusive soil properties are to be examined. It is demonstrated that, because determination of flow paths can be treated as a routing problem solved by applying a few simple rules, dynamic programming analysis is not limited by complexity of pore structures.     

On the manifestation of solonetzic properties in soils of the steppe and dry steppe zones in the south of Ukraine

Solonetzic soils are widespread in the south of Ukraine; some of them are represented by low-sodium solonetzes whose diagnostics is complicated by the absence of correspondence between the low content of exchangeable sodium and the clearly pronounced features of the illuvial solonetzic horizon. The statistical treatment of available literature, personal materials, and the materials of soil surveys in the south of Ukraine demonstrated that the diagnostics of low-sodium solonetzes should be based on the assessment of the degree of illuviation in the soil profile. It was shown that the degree of correlation between the content of exchangeable sodium and the degree of illuviation of the soil profile varies in dependence on the size of the statistical sample and differs for the low-sodium and medium-sodium solonetzes. Two scales characterizing the degree of soil illuviation were developed for these soil groups. It is suggested that these scales should be applied during soil surveys in the areas with solonetzic soils.     

An improved technique for agricultural implement draught analysis

Agricultural implement draft requirements show considerable spatial variability due to variations in soil properties and fracturing of soil. A large sample size is necessary to obtain a representative mean draft value for a given soil type and condition because of this variability. Moreover, empirical polynomial/multi-linear regression models for implement draft are often subjected to multi-collinearity problems. Proper design of experiments can assist in the complete elimination of these multicollinearity problems. An implement test procedure has been developed which addresses the problems of soil variability and multi-collinearity. Proper choice of values for independent variables in the experimental design phase can assist in transforming these variables to an orthogonal domain which completely overcomes multi-collinearity problems. The orthogonal regression technique using transformed variables and the conventional polynomial/multi-linear regression techniques using real variables were used to analyze draft data for a moldboard plow in a Capay clay soil to illustrate advantages of the orthogonal technique.     

引入时相信息的耕地土壤有机质遥感反演模型

土壤有机质(soil organic matter,SOM)是土壤质量评价的重要指标。监测SOM含量及其空间分布对土壤利用与保护、土壤有机碳库估算等具有重要意义。该文以松嫩平原典型区为研究区,采集4种主要土壤类型样本共147个,获取裸土期多时相MODIS地表反射率8 d合成产品,以单期、多期影像所构建光谱指数作为输入量,构建包含含水量变化与有机质含量信息的多光谱指数,建立SOM线性回归遥感反演模型,揭示SOM空间分布规律。结果表明:由于土壤含水量空间差异随时间变化,基于单期影像构建的模型主要输入量发生规律性改变,其中年积日137 d裸土条件最好,反演模型最优;比值光谱指数R61与SOM显著相关,而和含水量相关性极小,适于作为反演模型输入量;基于多期影像构建的模型引入时相信息后,精度与稳定性较单期影像模型显著提高,其中基于年积日137、105 d两期影像光谱指数所建立的多元线性模型最优;松嫩平原SOM呈现由东北向西南递减趋势。     

长期垄作免耕对不同大小土壤团聚体中几种氮素形态分布的影响

在西南大学国家紫色土肥力定位监测点,研究了长期（18年）垄作免耕和常规轮作两种耕作方式下,不同大小土壤团聚体中几种形态氮素含量及脲酶活性分布模式。结果表明,两种轮作方式下全氮、微生物体氮、NO3--N含量及脲酶活性在水稳性土壤团聚体中具有相似的分布模式。全氮和微生物体氮含量主要分布在 0.25 mm粒级的大团聚体中,其次是0.25~0.053 mm粒级的微团聚体中,粉砂与粘粒组分(0.053 mm)中分布最少;2.0~0.25 mm粒级团聚体中脲酶活性最高,0.25~0.053 mm的微团聚体中脲酶活性最低;NH4+-N和NO3-N在2.0 mm粒级团聚体中的含量均比其它粒级高。垄作免耕下各级团聚体中全氮、微生物体氮、NO3--N含量及脲酶活性均高于常规轮作。其中,2.0~0.25 mm粒级团聚体中的全氮含量高23.1%;2.0 mm粒级团聚体中NO3--N含量高26.7％。表明垄作免耕下各团聚体中全氮、微生物氮含量及脲酶活性显著高于常规轮作,但氮素含量及脲酶活性在不同粒径大小的土壤水稳性团聚体中的分布模式决定于土壤结构体本身,耕作方式的影响不显著。     

Maps of averaged spectral deviations from soil lines and their comparison with traditional soil maps

The analysis of 34 cloudless fragments of Landsat 5, 7, and 8 images (1985–2014) on the territory of Plavsk, Arsen’evsk, and Chern districts of Tula oblast has been performed. It is shown that bare soil surface on the RED–NIR plots derived from the images cannot be described in the form of a sector of spectral plane as it can be done for the NDVI values. The notion of spectral neighborhood of soil line (SNSL) is suggested. It is defined as the sum of points of the RED–NIR spectral space, which are characterized by spectral characteristics of the bare soil applied for constructing soil lines. The way of the SNSL separation along the line of the lowest concentration density of points on the RED–NIR spectral space is suggested. This line separates bare soil surface from vegetating plants. The SNSL has been applied to construct soil line (SL) for each of the 34 images and to delineate bare soil surface on them. Distances from the points with averaged RED–NIR coordinates to the SL have been calculated using the method of moving window. These distances can be referred to as averaged spectral deviations (ASDs). The calculations have been performed strictly for the SNSL areas. As a result, 34 maps of ASDs have been created. These maps contain ASD values for 6036 points of a grid used in the study. Then, the integral map of normalized ASD values has been built with due account for the number of points participating in the calculation (i.e., lying in the SNSL) within the moving window. The integral map of ASD values has been compared with four traditional soil maps on the studied territory. It is shown that this integral map can be interpreted in terms of soil taxa: the areas of seven soil subtypes (soddy moderately podzolic, soddy slightly podzolic, light gray forest. gray forest, dark gray forest, podzolized chernozems, and leached chernozems) belonging to three soil types (soddy-podzolic, gray forest, and chernozemic soils) can be delineated on it.     

基于无人机图像和贝叶斯CSRNet模型的粘连云杉计数

自动、准确且快速地统计苗木数量是实现苗圃高效管理的重要基础。针对现有苗木计数方法准确率较低且无法准确统计粘连苗木等问题,该研究提出了一种基于贝叶斯CSRNet模型的云杉计数模型。该模型以对粘连苗木具有良好稳定性的CSRNet模型为基础,引入贝叶斯损失函数,以人工标注的点标签数据作为监督信号。以1 176幅云杉图像训练贝叶斯CSRNet模型,并通过166幅测试集云杉图像测试。结果表明,贝叶斯CSRNet模型可以准确、快速地统计无人机航拍图像内的云杉,对测试集图像内云杉的平均计数准确率（Mean Counting Accuracy,MCA）、平均绝对误差（Mean Absolute Error,MAE）和均方误差（Mean Square Error,MSE）分别为99.19%、1.42和2.80。单幅云杉图像耗时仅为248 ms,模型大小为62 Mb。对比YOLOv3模型、改进YOLOv3模型、CSRNet模型和贝叶斯CSRNet模型对166幅测试集云杉图像的计数结果,贝叶斯CSRNet模型的MCA分别比YOLOv3模型、改进YOLOv3模型、CSRNet模型高3.43%、1.44%和1.13%;贝叶斯CSRNet模型的MAE分别比YOLOv3模型、改进YOLOv3模型、CSRNet模型低6.8、2.9和1.67;贝叶斯CSRNet模型的MSE分别比YOLOv3模型、改进YOLOv3模型、CSRNet模型低101.74、23.48和8.57。在MCT和MS两项指标上,贝叶斯CSRNet模型与CSRNet模型相同且优于YOLOv3模型和改进YOLOv3模型。贝叶斯CSRNet模型可实现无人机航拍图像内苗木数量的自动、准确、快速统计,为苗木库存智能盘点提供参考。     

利用土壤的序列数字图像技术研究孔隙小尺度特征

将土壤样品用混有荧光有机染料的树脂浸渍后 ,以 1mm间距连续研磨 ,经过一定的图像转换处理 ,最后以连续数字图像方式对图像孔隙的小尺度特征进行分析。对一种美国耕作黑土和一种法国耕作黄壤的比较研究表明 ,两类土壤在图像孔隙面积、图像孔隙面积变异性、图像孔隙孔径、孔隙连接度、孔隙重叠度以及孔隙流通性等方面均有所差异 ,这与两类土壤不同的有机质含量有关。     

Traffic and seasonal influences on the energy required for cultivation and on the subsequent tilth

Two similar experiments were conducted over a 3 year period on a clay soil to compare the effects of different intensities of traffic, ranging from 0 to 124 t km ha⁻¹ (1.24 MN km ha⁻¹) on the energy required for ploughing and secondary cultivation. The tilths produced by these operations were measured by sieving and, in one year, by image analysis of sections cut from large blocks of resin-impregnated soil.

Both energy required for cultivation and mean aggregate size after ploughing were related to the intensity of wheeling and the soil water content at ploughing. In the absence of wheeling, energy for ploughing was reduced by between 17 and 45%, and mean aggregate size was reduced to 20–70% of that resulting from a conventional tractor and plough-based system.

Specific draught in the driest soil conditions was almost twice that in the wettest conditions. The geometric mean diameter of aggregates produced by the mouldboard plough ranged from 7 mm in a wetter year and with no traffic to 140 mm in the driest year with a year-round tractor traffic system.

Scanning sections of large impregnated blocks with an image analyser provided more detailed information on aggregate size distributions within the plough layer than did sieving, and it also provided information on pore size distributions. This technique served to illustrate large differences in soil macrostructure between a conventional tractor and plough cultivation system compared with an untrafficked but ploughed soil.