基于改进YOLO v5的夜间温室番茄果实快速识别

为实现日光温室夜间环境下采摘机器人正常工作以及番茄快速识别,提出一种基于改进YOLO v5的夜间番茄果实的识别方法。采集夜间环境下番茄图像2 000幅作为训练样本,通过建立一种基于交并比的CIOU目标位置损失函数,对原损失函数进行改进,根据计算函数anchor生成自适应锚定框,确定最佳锚定框尺寸,构建改进型YOLO v5网络模型。试验结果表明,改进YOLO v5网络模型对夜间环境下番茄绿色果实识别精度、红色果实识别精度、综合平均识别精度分别为96.2%、97.6%和96.8%,对比CNN卷积网络模型及YOLO v5模型,提高了被遮挡特征物与暗光下特征物的识别精度,改善了模型鲁棒性。将改进YOLO v5网络模型通过编译将训练结果写入安卓系统制作快速检测应用软件,验证了模型对夜间环境下番茄果实识别的可靠性与准确性,可为番茄实时检测系统的相关研究提供参考。     

基于改进YOLO v5s的轻量化植物识别模型研究

为方便调查宁夏全区荒漠草原植物种类及其分布,需对植物识别方法进行研究。针对YOLO v5s模型参数量大,对复杂背景下的植物不易识别等问题,提出一种复杂背景下植物目标识别轻量化模型YOLO v5s-CBD。改进模型YOLO v5s-CBD在特征提取网络中引入带有Transformer模块的主干网络BoTNet(Bottleneck transformer network),使卷积和自注意力相结合,提高模型的感受野;同时在特征提取网络融入坐标注意力(Coordinate attention, CA),有效捕获通道和位置的关系,提高模型的特征提取能力;引入SIoU函数计算回归损失,解决预测框与真实框不匹配问题;使用深度可分离卷积(Depthwise separable convolution, DSC)减小模型内存占用量。实验结果表明,YOLO v5s-CBD模型在单块Nvidia GTX A5000 GPU单幅图像推理时间仅为8 ms,模型内存占用量为8.9 MB,精确率P为95.1%,召回率R为92.9%,综合评价指标F1值为94.0%,平均精度均值(mAP)为95.7%,在VOC数据集...     

基于注意力机制与改进YOLO的温室番茄快速识别

为了实现复杂环境下农业机器人对番茄果实的快速准确识别,提出了一种基于注意力机制与改进YOLO v5s的温室番茄目标快速检测方法。根据YOLO v5s模型小、速度快等特点,在骨干网络中加入卷积注意力模块(CBAM),通过串联空间注意力模块和通道注意力模块,对绿色番茄目标特征给予更多的关注,提高识别精度,解决绿色番茄在相似颜色背景中难识别问题;通过将CIoU Loss替换GIoU Loss作为算法的损失函数,在提高边界框回归速率的同时提高果实目标定位精度。试验结果表明,CB-YOLO网络模型对温室环境下红色番茄检测精度、绿色番茄检测精度、平均精度均值分别为99.88%、99.18%和99.53%,果实检测精度和平均精度均值高于Faster R-CNN模型、YOLO v4-tiny模型和YOLO v5模型。将CB-YOLO模型部署到安卓手机端,通过不同型号手机测试,验证了模型在移动终端设备上运行的稳定性,可为设施环境下基于移动边缘计算的机器人目标识别及采收作业提供技术支持。     

基于坐标注意力机制和YOLO v5s模型的山羊脸部检测方法

山羊的脸部检测对羊场的智能化管理有着重要的意义。针对实际饲养环境中,羊群存在多角度、分布随机、灵活多变、羊脸检测难度大的问题,以YOLO v5s为基础目标检测网络,提出了一种结合坐标信息的山羊脸部检测模型。首先,通过移动设备获取舍内、舍外、单头以及多头山羊的图像并构建数据集。其次,在YOLO v5s的主干网络融入坐标注意力机制,以充分利用目标的位置信息,提高遮挡区域、小目标、多视角样本的检测精度。试验结果表明,改进YOLO v5s模型的检测精确率为95.6%,召回率为83.0%,mAP_0.5为90.2%,帧速率为69 f/s,模型内存占用量为13.2 MB;与YOLO v5s模型相比,检测精度提高1.3个百分点,模型所占内存空间减少1.2 MB;且模型的整体性能远优于Faster R-CNN、YOLO v4、YOLO v5s模型。此外,本文构建了不同光照和相机抖动的数据集,来进一步验证本文方法的可行性。改进后的模型可快速有效地对复杂场景下山羊的脸部进行精准检测及定位,为动物精细化养殖时目标检测识别提供了检测思路和技术支持。     

基于改进YOLO v5s的轻量级奶牛体况评分方法

奶牛体况评分是评价奶牛产能与体态健康的重要指标。目前,随着现代化牧场的发展,智能检测技术已被应用于奶牛精准养殖中。针对目前检测算法的参数量多、计算量大等问题,以YOLO v5s为基础,提出了一种改进的轻量级奶牛体况评分模型(YOLO-MCE)。首先,通过2D摄像机在奶牛挤奶通道处采集奶牛尾部图像并构建奶牛BCS数据集。其次,在MobileNetV3网络中融入坐标注意力机制(Coordinate attention, CA)构建M3CA网络。将YOLO v5s的主干网络替换为M3CA网络,在降低模型复杂度的同时,使得网络特征提取时更关注于牛尾区域的位置和空间信息,从而提高了运动模糊场景下的检测精度。YOLO v5s预测层采用EIoU Loss损失函数,优化了目标边界框回归收敛速度,生成定位精准的预测边界框,进而提高了模型检测精度。试验结果表明,改进的YOLO v5s模型的检测精度为93.4%,召回率为85.5%,mAP@0.5为91.4%,计算量为2.0×10⁹,模型内存占用量仅为2.28 MB。相较原始YOLO v5s模型,其计算量降低87.3%,模型内存占用量减...     

基于YOLO v5-TL的褐菇采摘视觉识别-测量-定位技术

为实现褐菇高效、精准、快速的自动化采摘,针对工厂化褐菇的种植特点,提出一种基于YOLO v5迁移学习(YOLO v5-TL)结合褐菇三维边缘信息直径动态估测法的褐菇原位识别-测量-定位一体化方法。首先,基于YOLO v5-TL算法实现复杂菌丝背景下的褐菇快速识别;再针对锚框区域褐菇图像进行图像增强、去噪、自适应二值化、形态学处理、轮廓拟合进行褐菇边缘定位,并提取边缘点和褐菇中心点的像素坐标;最后基于褐菇三维边缘信息的直径动态估测法实现褐菇尺寸的精确测量和中心点定位。试验结果表明单帧图像平均处理时间为50ms,光照强度低、中、高情况下采摘对象识别平均成功率为91.67%,其中高光强时识别率达100%,菇盖的尺寸测量平均精度为97.28%。研究表明,本文提出的YOLO v5-TL结合褐菇三维边缘信息直径动态估测法可实现工厂化种植环境下褐菇识别、测量、定位一体化,满足机器人褐菇自动化采摘需求。     

基于改进YOLO v3模型的奶牛发情行为识别研究

为提高复杂环境下奶牛发情行为识别精度和速度,提出了一种基于改进YOLO v3模型的奶牛发情行为识别方法。针对YOLO v3模型原锚点框尺寸不适用于奶牛数据集的问题,对奶牛数据集进行聚类,并对获得的新锚点框尺寸进行优化;针对因数据集中奶牛个体偏大等原因而导致模型识别准确率低的问题,引入DenseBlock结构对YOLO v3模型原特征提取网络进行改进,提高了模型识别性能;将YOLO v3模型原边界框损失函数使用均方差(MSE)作为损失函数度量改为使用F_IoU和两框中心距离D_c度量,提出了新的边界框损失函数,使其具有尺度不变性。从96段具有发情爬跨行为的视频片段中各选取50帧图像,根据发情爬跨行为在活动区出现位置的不确定性和活动区光照变化的特点,对图像进行水平翻转、±15°旋转、随机亮度增强(降低)等数据增强操作,用增强后的数据构建训练集和验证集,对改进后的模型进行训练,并依据F1、mAP、准确率P和召回率R指标进行模型优选。在测试集上的试验表明,本文方法模型的识别准确率为99.15%,召回率为97.62%,且处理速度达到31 f/s,能够满足复...     

基于改进YOLO v5n的舍养绵羊行为识别方法

日常行为是家畜健康状况的重要体现,在传统的行为识别方法中,通常需要人工或者依赖工具对家畜进行观察。为解决以上问题,基于YOLO v5n模型,提出了一种高效的绵羊行为识别方法,利用目标识别算法从羊圈斜上方的视频序列中识别舍养绵羊的进食、躺卧以及站立行为。首先用摄像头采集养殖场中羊群的日常行为图像,构建绵羊行为数据集;其次在YOLO v5n的主干特征提取网络中引入SE注意力机制,增强全局信息交互能力和表达能力,提高检测性能;采用GIoU损失函数,减少训练模型时的计算开销并提升模型收敛速度;最后,在Backbone主干网络中引入GhostConv卷积,有效地减少了模型计算量和参数量。实验结果表明,本研究提出的GS-YOLO v5n目标检测方法参数量仅为1.52×106,相较于原始模型YOLO v5n减少15%;浮点运算量为3.3×109,相较于原始模型减少30%;且平均精度均值达到95.8%,相比于原始模型提高4.6个百分点。改进后模型与当前主流的YOLO系列目标检测模型相比,在大幅减少模型计算量和参数量的同时,检测精度均有较高提升。在边缘设备上进行部署,达到了实时检测要求,可准确快速地对绵羊进行定位并检测。     

基于YOLO v5与短时跟踪的鸡只呼吸道疾病早期检测

针对鸡只呼吸困难这一早期呼吸道疾病显著症状难以检测的问题,提出一种基于YOLO v5与短时跟踪的鸡只呼吸道疾病早期检测方法。对YOLO v5算法进行锚框自适应设置与CIoU Loss (Complete IoU Loss)应用等特定优化后,用于群鸡复杂环境中准确识别鸡头目标并检测是否为张口状态。根据鸡头坐标框交并比实现鸡头目标短时跟踪并获取不同鸡头的短时动作序列,再对动作序列进行分析,判断张口-闭口组合出现的频率,动态检测是否存在鸡只呼吸困难情况。实验结果表明,改进YOLO v5算法检测鸡头目标的mAP为80.1%,张口检测准确率为67.3%,闭口检测准确率为92.8%,基于时间序列的呼吸困难行为检测方法的识别准确率为91.8%,召回率为75%,精准率为67.9%,可为群鸡养殖环境中的鸡只早期呼吸道疾病检测提供参考。     

基于改进CenterNet的玉米雄蕊无人机遥感图像识别

为准确识别抽雄期玉米雄蕊实现监测玉米长势、植株计数和估产,基于无锚框的CenterNet目标检测模型,通过分析玉米雄蕊的尺寸分布,并在特征提取网络中添加位置坐标,从而提出一种改进的玉米雄蕊识别模型。针对雄蕊尺寸较小的特点,去除CenterNet网络中对图像尺度缩小的特征提取模块,在降低模型参数的同时,提高检测速度。在CenterNet特征提取模型中添加位置信息,提高定位精度,降低雄蕊漏检率。试验结果表明,与有锚框的YOLO v4、Faster R-CNN模型相比,改进的CenterNet雄蕊检测模型对无人机遥感影像的玉米雄蕊识别精度达到92.4%,分别高于Faster R-CNN和YOLO v4模型26.22、3.42个百分点;检测速度为36f/s,分别比Faster R-CNN和YOLO v4模型高32、23f/s。本文方法能够准确地检测无人机遥感图像中尺寸较小的玉米雄蕊,为玉米抽雄期的农情监测提供参考。     

Overland water and salt flows in a set of rice paddies

Cultivation of paddy rice in semiarid areas of the world faces problems related to water scarcity. This paper aims at characterizing water use in a set of paddies located in the central Ebro basin of Spain using experimentation and computer simulation. A commercial field with six interconnected paddies, with a total area of 5.31 ha, was instrumented to measure discharge and water quality at the inflow and at the runoff outlet. The soil was classified as a Typic Calcixerept, and was characterized by a mild salinity (2.5 dS m⁻¹) and an infiltration rate of 5.8 mm day⁻¹. The evolution of flow depth at all paddies was recorded. Data from the 2002 rice-growing season was elaborated using a mass balance approach to estimate the infiltration rate and the evolution of discharge between paddies. Seasonal crop evapotranspiration, estimated with the surface renewal method, was 731 mm (5.1 mm day⁻¹), very similar to that of other summer cereals grown in the area, like corn. The irrigation input was 1874 mm, deep percolation was 830 mm and surface runoff was 372 mm. Irrigation efficiency was estimated as 41%. The quality of surface runoff water was slightly degraded due to evapoconcentration and to the contact with the soil. During the period 2001–2003, the electrical conductivity of surface runoff water was 54% higher than that of irrigation water. However, the runoff water was suitable for irrigation. A mechanistic mass balance model of inter-paddy water flow permitted to conclude that improvements in irrigation efficiency cannot be easily obtained in the experimental conditions. Since deep percolation losses more than double surface runoff losses, a reduction in irrigation discharge would not have much room for efficiency improvement. Simulations also showed that rice irrigation performance was not negatively affected by the fluctuating inflow hydrograph. These hydrographs are typical of turnouts located at the tail end of tertiary irrigation ditches. In fact, these are the sites where rice has been historically cultivated in the study area, since local soils are often saline-sodic and can only grow paddy rice taking advantage of the low salinity of the irrigation water. The low infiltration rate characteristic of these saline-sodic soils (an experimental value of 3.2 mm day⁻¹ was obtained) combined with a reduced irrigation discharge resulted in a simulated irrigation efficiency of 60%. Paddy rice irrigation efficiency can attain reasonable values in the local saline-sodic soils, where the infiltration rate is clearly smaller than the average daily rice evapotranspiration.     

瓦房店市水果生产标准化发展研究

瓦房店市是我国重要的水果生产基地,在国内水果生产领域占有重要地位。随着水果市场化的不断深入,如何满足和适应国内外市场需求成为瓦房店市水果生产的关键问题。从瓦房店市水果生产实际出发,阐述了实施水果生产标准化的必要性,提出了发展水果生产标准化的具体措施。     

金谷子耐旱TG118引种试种及推广应用

对引进的谷子新品种金谷子耐旱TG118进行品种比较试验和多点试种。结果表明：该品种抗旱、耐瘠、适应性强,稳产、高产、抗病抗风抗倒伏强,品质优良,粮草兼用,种植简单,省工易管;一般产量在6000kg/hm2以上,比当地主栽品种增产20%以上。最高产量达9450kg/hm2,适于在辽西北地区推广应用。     

对《种子法》存在问题的探讨

《种子法》是我国种植业第一部法律。自实施以来,为依法加强种子管理提供了有力的法律支撑,促进我国种子产业发生翻天覆地的变化。但随着依法管理工作的深入和种子产业的快速发展,《种子法》及配套规章的一些问题逐步显现出来。对《种子法》及配套规章存在的问题进行深入分析,对需要进一步补充的章节以及需要调整的内容进行深入探讨,以求更加完善和明确。     

Review of the Egyptian experience in implementing land drainage projects

About 30 years have passed since the Government of Egypt embarked on implementing a series of large scale drainage projects. At present, about 3.8 million acres have been provided with drainage systems on the basis of systematic pre-drainage investigations and designs. The target is to provide drains in approximately 6.4 million acres in the Nile Valley and Delta.The implementation of the subsurface drainage system is carried out by the public sector and private contractors under direct supervision of governmental regional departments. The implementation process depends on many factors related to the drainage material, machinery, manpower, site requirements, farmers and organizations involved. Problems and constraints are sometimes challenging, however, the annual rate of implementation has gradually increased to 170,000 acre/year.This paper discusses the different aspects involved in the implementation process of drainage systems. The development in materials, machinery and construction technologies will be reviewed. Institutional and management factors are going to be also considered.     

和谐社会理论建设的几点哲学思考

在新世纪,面对机遇和挑战,党的十六大把“社会更加和谐”作为全面建设更高水平的小康社会的一个重要目标。中国人素有社会和谐的理想;西方人自古也向往社会和谐;借鉴东西方两种和谐社会思想的经验、教训,构建社会主义和谐社会理论。     

A European irrigation map for spatially distributed agricultural modelling

We present a pan-European irrigation map based on regional European statistics, a European land use map and a global irrigation map. The map provides spatial information on the distribution of irrigated areas per crop type which allows determining irrigated areas at the level of spatial modelling units. The map is a requirement for a European scale assessment of the impacts of irrigated agriculture on water resources based on spatially distributed modelling of crop growth and water balance. The irrigation map was compiled in a two step procedure. First, irrigated areas were distributed to potentially irrigated crops at a regional level (European statistical regions NUTS3), combining Farm Structure Survey (FSS) data on irrigated area, crop-specific irrigated area for crops whenever available, and total crop area. Second, crop-specific irrigated area was distributed within each statistical region based on the crop distribution given in our land use map. A global map of irrigated areas with a 5′ resolution was used to further constrain the distribution within each NUTS3 based on the density of irrigated areas. The constrained distribution of irrigated areas as taken from statistics to a high resolution dataset enables us to estimate irrigated areas for various spatial entities, including administrative, natural and artificial units, providing a reasonable input scenario for large-scale distributed modelling applications. The dataset bridges a gap between global datasets and detailed regional data on the distribution of irrigated areas and provides information for various assessments and modelling applications.     

粉质仪和拉伸仪在面粉生产中的应用

在介绍粉质仪和拉伸仪结构和原理的基础上,探讨了粉质曲线中吸水率、形成时间、稳定时间、弱化度和评价值等指标与面粉品质的关系,以及如何根据拉伸曲线中面团最大拉伸阻力、延伸度和拉伸曲线面积等指标来评价面团品质,并结合生产实际,概述了粉质仪和拉伸仪在面粉生产中的作用。     

农村信息化服务的现状及建设措施

分析当前我国农村信息化服务的现状,提出加强农村信息化服务的新举措。     

车辆ABS控制算法的研究及探讨

对现代汽车的ABS控制算法作了一些简单的介绍,同时指出了它们的一些缺陷,相应地提出了几种对ABS的其它控制方法并加以探讨,其中着重论述了ABS门限值控制算法中参考车速的确定方法和基于附着系数的控制算法。