基于机器视觉的猪体质量估测模型比较与优化

基于机器视觉的猪体质量估测模型较多,但模型缺乏在实用性、准确性的对比,最佳模型没有定论。该文总结了已有的估测算法,基于79组背部图像面积、实际面积、体长、体宽、体高、臀宽、臀高数据,使用线性回归、幂回归、二次回归、主成分线性回归、RBF(radial basis function,径向基函数)神经网络等方法,重建了13种体质量估测模型,并比较了13种模型的估测精度。结果表明,基于体长、体宽、体高、臀宽和臀高的线性回归模型具有较好的估测精度,估测值与真值的相关系数达到了0.996。利用主成分法去掉体尺的共线性,利用曲线回归解决残差不均匀问题,更加符合猪体质量增长趋势,结果表明基于主成分的幂回归模型具有较高的相关系数和较低的标准估计误差,对于97组数据的估测平均相对误差为2.02%。使用猪场实测24组数据验证模型,估测质量与测量值相关系数为0.97,估测平均相对误差为2.26%,标准差为1.78%,优于基于面积和面积体高结合的估测模型,平均绝对误差为2.08 kg,优于面积体高结合方法的平均绝对误差。试验证明使用多个体尺的主成分幂回归体质量估测模型较为精确,可用于机器视觉估测猪体质量的应用中。     

生猪肉产量预测的非接触实时在线机器视觉系统

为了准确估测生猪肉产量,该文提出了一种自动提取生猪活体特征的方法,建立了生猪肉产量预测算法。首先,基于自主研发的生猪肉产量预测的机器视觉系统装置实时采集54头活体大白猪的俯视图和侧视图,然后根据图像处理技术对所得的图像进行分析,提取生猪的体长、胸宽、臀宽、体高等体尺参数值;同时通过称量装置获得每头猪的活体质量。其次,将54头生猪样品按2∶1的比例分为校正集和验证集,基于体长、胸宽、臀宽、体高和体质量5个特征参数分别利用多元线性回归法(multiple linear regression,MLR)和偏最小二乘回归法(partial least-squares regression,PLSR)建立不同的肉产量预测模型,通过比较各预测模型的估测精度、相关系数等,确定最佳预测模型。分析结果表明,5个参数中体质量对肉产量预测的权重最大;偏最小二乘回归方法的预测结果较好,最佳预测模型的预测相关系数为0.95,预测误差为3.09 kg。试验结果证实,在检测系统中使用基于生猪的体尺参数和体质量建立的偏最小二乘回归预测模型可快速准确预测生猪肉产量,适于实际生产应用。     

基于图像的肉羊生长参数实时无接触监测方法

针对基于体尺、体质量的肉羊实时生长监测中体尺、体质量需要人工测量的不足,论文提出基于结构化限位装置及机器视觉技术的无接触肉羊生长参数测量方法,并讨论无接触方法获取的体尺数据与羊只生长特性的关系。首先,基于自主研发的无应激形态参数采集系统实时采集60只小尾寒羊的俯视图和侧视图,应用图像处理技术对所得到的图像进行分析,提取体高、臀高、体长、胸深、胸宽、臀宽3类6种体尺参数;无接触生长参数采集系统同时记录对应羊只的体质量。对无接触方法获取的体尺、体质量数据相关关系进行研究,并分别利用单因素线性回归、单因素非线性回归、多元线性回归、偏最小二乘回归、RBF网络拟合、SVM回归方法建立体尺与体质量关系模型。试验表明:体高、臀高、体长、胸深、胸宽、臀宽的最大相对误差分别为4.73%、2.55%、2.50%、3.95%、3.80%和2.90%;无接触方法获取的体高、体长、胸宽与体质量相关性大于0.8;在基于单因素的生长监测中可选择体长参数;多体尺能够较全面地表达羊只的生长状态,其中胸深、胸宽、体长是重要的监测参数;多因素非线性模型可以更全面、精准的体现羊只生长特性。论文提出的无接触方法可有效提升工作效率,节约50%的人工投入。同时,也可减少羊只的应激反应,是长期、实时监测羊只生长的实用方法,对推动精准、福利化养羊具有重要意义。     

基于改进Faster R-CNN识别深度视频图像哺乳母猪姿态

猪舍场景下,昼夜交替光线变化、热灯光照影响,及仔猪与母猪的粘连等因素,给全天候哺乳母猪姿态自动识别带来很大困难。该文以深度视频图像为数据源,提出基于改进Faster R-CNN的哺乳母猪姿态识别算法。将残差结构引入ZF网络,设计ZF-D2R网络,以提高识别精度并保持实时性;将Center Loss监督信号引入Faster R-CNN训练中,以增强类内特征的内聚性,提升识别精度。对28栏猪的视频图像抽取站立、坐立、俯卧、腹卧和侧卧5类姿态共计7 541张图像作为训练集,另取5类姿态的5 000张图像作为测试集。该文提出的改进模型在测试集上对哺乳母猪的站立、坐立、俯卧、腹卧和侧卧5类姿态的识别平均准确率分别达到96.73%、94.62%、86.28%、89.57%和99.04%,5类姿态的平均准确率均值达到93.25%。在识别精度上,比ZF网络和层数更深的VGG16网络的平均准确率均值分别提高了3.86和1.24个百分点。识别速度为0.058 s/帧,比VGG16网络速度提高了0.034 s。该文方法在提高识别精度的同时保证了实时性,可为全天候母猪行为识别提供技术参考。     

广西白猪活体可观测性状与屠体瘦肉率的相关分析

本文采用逐步回归分析法分析广西白猪的体长、胸围、体长与胸围的比率、背膘厚、体重等活体可观测的性状与屠体瘦肉率的相关性,建立回归方程: Y=34.6 29.17x_3-3.14x_4 作F检脸结果,回归关系高度显著。用此回归方程估测瘦肉率,49头猪的实测瘦肉率没有一头超出Y士2se的范围。据此回归方程的估测结果,确定广西白猪的选种标准为:对满六月龄的后备公、母猪进行活体测定,凡是体长与胸围的比率达到1.15以上,背膘厚在3cm以下者选入育种核心群。从第四世代起执行这个选择方案,经过第四、第五、第六世代三个世代的选择,使90公斤左右育肥猪的瘦肉率从55.46％提高到56.71％。选择效果是令人满意的。     

基于Kinect传感器的无接触式母猪体况评分方法

为了提高母猪的繁殖性能,减少传统方法给动物和估测人员带来的不利影响,该研究提出了一种可应用于实际生产中的准确、无接触式的母猪体况评分(body condition scoring,BCS)方法。试验使用Kinect传感器获取108组母猪臀股的三维图像,选取48组图像进行分析处理,计算出臀部的高宽比、臀股面积及曲率半径。试验结果表明:母猪臀部的高宽比、臀股面积和曲率半径与背膘厚度的相关系数分别为0.567、0.502、0.951;以曲率半径作为主要参数建立母猪体况估测模型。取剩余60组图像进行验证,估测模型计算结果与经验方法评估结果差异较小,准确率达到91.7%;结果表明,基于三维重构技术的Kinect传感器能够实现母猪在饲养管理过程中对体况的无接触式检测。     

用于定向播种的玉米种穗图像精选方法

玉米定向播种,要求籽粒形状扁平、具有方向性。为了减少玉米粒精选的工作量,该文以玉米种穗为对象,研究适合定向播种的玉米种穗图像精选方法。设计玉米种穗精选传输装置,实现了对玉米种穗的动态图像采集和精选。根据种穗的外形特征判断小种穗;利用R+G-2B方法加强黄色籽粒区域,根据黄色籽粒区域与整个种穗的面积比判断缺粒及霉变种穗;利用种穗图像的横向和纵向像素值累计分布特征,追踪中间穗行的籽粒轮廓,并通过其端面矩形度,判断籽粒合格的种穗。随机抽检50个种穗样本,结果表明:外形特征的检测准确率为100%,缺粒及霉变的检测准确率为96%,穗上籽粒端面矩形度的检测准确率为98%,总体检测准确率为94%。该文为定向播种用玉米籽粒精选前期的种穗精选提供了一种图像识别方法。     

基于射线轮廓点匹配的生猪红外与可见光图像自动配准

为研究生猪多源图像特征提取方法及生猪体表温度与生猪异常的关系特征,该文提出一种基于射线轮廓特征点匹配的红外与可见光图像自动配准方法。采用红外热像仪,同时采集相同猪舍场景的可见光图像和红外热图像,以红外热图像中生猪区域质心为中心间隔均匀角度构建辅助射线,提取射线与边缘轮廓交点构建匹配特征点集,通过计算不同尺度变换因子下特征点集间的加权部分Hausdorff距离作为测度,引入RPROP算法进行迭代加速,实现了可见光图像和红外热图像的快速自动配准。试验中,应用该文算法对50对红外和光学图像进行了测试,所提出自动配准方法配准成功率达到94%,平均配准误差小于1像素,试验结果表明自动配准效果达到或超过手动配准的效果,为进一步研究生猪多源图像异常特征提取奠定基础。     

基于深度和彩色双信息特征源的Kinect植物图像拼接

图像拼接对制作全景图具有重要作用,传统SIFT(scale-invariant feature transform)图像拼接方法受光照不均匀或风吹影响,存在错位和缺失等情况。针对这一问题,该文提出一种基于Kinect彩色和深度双信息特征源的图像拼接方法。首先对获取的植株彩色图像采用SIFT算法进行特征点提取和特征点匹配,利用Kinect收集到的植株深度数据去除误匹配,采用RANSAC算法寻找投影变换矩阵,最后通过最佳缝合线算法进行图像融合。室内和室外试验结果表明,该文基于Kinect彩色和深度双信息特征源的图像拼接方法能够有效克服光照、风吹等环境因素的影响,避免了图像缺失、亮暗差异、重影等拼接错误,该文方法图像拼接时间较短,平均匹配准确率达96.0%,较传统SIFT图像拼接方法平均准确率提高了5.9%。     

多视角深度相机的猪体三维点云重构及体尺测量

对活体牲畜三维重构,数据采集方式、快速配准融合方法、表型体尺测量方法缺乏成熟有效的方案,导致目前活体牲畜的自动体尺测量技术难以在养殖场中推广应用。该文以猪为研究对象运用消费级深度相机KinectV2从正上方和左右两侧3个不同角度同步获取在采集通道中自由行走猪的局部点云。局部点云采用邻域曲率变化法去噪,并运用基于轮廓连贯性点云配准融合,最后采用多体尺数据精确估算技术测定包括体长、体高、胸宽、腹围等数据。该文分别对比实验室中模型猪由传输带以5种不同速率经过通道和养殖场内25头猪逐一经过通道,2种情况下采集数据进行各项体尺测算结果。其结果显示模型猪在传输带上以0、0.3、0.6、0.9和1.2 m/s等5种不同速率下测量体长、体高、胸宽、腹围值与实测值的平均相对误差分别为1.77%、1.36%、2.74%和2.17%。养殖环境下对25头猪同样4种体尺值与实测值的平均相对误差分别为2.56%,2.32%,3.89%和4.51%。试验结果发现养殖场活体猪测量最小误差可以达到实验室环境下的效果,但是最大相对误差变化较大,其原因在于养殖场中猪自由行走采集数据时行为姿态发生很大变化。     

Nitrification in soils incubated with pig slurry

Incubation studies (5 weeks at 30°C) of nitrification were made in an acid (pH 5.8) and a neutral (pH 7.1) soil receiving varying concentrations of pig slurry and (NH₄)₂SO₄ solution. Mineral-N and pH changes were observed at weekly intervals and inorganic salts media were used to obtain separate estimates of the numbers of NH₄-N- and NO₂-N-oxidizing bacteria. In the acid soil, pig slurry NH₄-N was nitrified to a greater extent than (NH₄)₂SO₄. In the neutral soil, slurry additions resulted in the accumulation of NO₂^?-N and, in one case, the complete inhibition of nitrification for 4 weeks. Slurry raised the pH of both soils more than (NH₄)₂SO₄ and nitrification in the acid soil was most rapid in a 2 week period of elevated pH following slurry applications. Numbers of Nitroxomonas isolated from the acid soil were considered high enough to account for NH₄-N oxidation in slurry-treated samples. Numbers of nitrifiers recovered from the incubated neutral soil samples were variable but frequently high enough (>10⁴/g dry soil) to account for observed rates of nitrification. Results are discussed in relation to heterotrophic nitrification in soils, and the practical implications of spreading slurry on agricultural land.     

Nitrogen losses from outdoor pig farming systems

Abstract. Nitrogen losses via nitrate leaching, ammonia volatilization and nitrous oxide emissions were measured from contrasting outdoor pig farming systems in a two year field study. Four 1‐ha paddocks representing three outdoor pig management systems and an arable control were established on a sandy loam soil in Berkshire, UK. The pig management systems represented: (i) current commercial practice (CCP) ‐ 25 dry sows ha^?1 on arable stubble; (ii) ‘improved’ management practice (IMP) ‐ 18 dry sows ha^?1 on stubble undersown with grass, and (iii) ‘best’ management practice (BMP) 12 dry sows ha^?1 on established grass. Nitrogen (N) inputs in the feed were measured and N offtakes in the pig meat estimated to calculate a nitrogen balance for each system. In the first winter, mean nitrate‐N concentrations in drainage water from the CCP, IMP, BMP and arable paddocks were 28, 25, 8 and 10 mg NO₃ l^?1, respectively. On the BMP system, leaching losses were limited by the grass cover, but this was destroyed by the pigs before the start of the second drainage season. In the second winter, mean concentrations increased to 111, 106 and 105 mg NO₃‐N l^?1 from the CCP, IMP and BMP systems, respectively, compared to only 32 mg NO₃‐N l^?1 on the arable paddock. Ammonia (NH₃) volatilization measurements indicated that losses from outdoor dry sows were in the region of 11 g NH₃‐N sow^?1 day^?1. Urine patches were identified as the major source of nitrous oxide (N₂O) emissions, with N₂O‐N losses estimated at less than 1% of the total N excreted. The nitrogen balance calculations indicated that N inputs to all the outdoor pig systems greatly exceeded N offtakes plus N losses, with estimated N surpluses on the CCP, IMP and BMP systems after 2 years of stocking at 576, 398 and 264 kg N ha^?1, respectively, compared with 27 kg N ha^?1 on the arable control. These large N surpluses are likely to exacerbate nitrate leaching losses in following seasons and make a contribution to the N requirement of future crops.     

舍饲条件下猪只福利的实现与设施配套

舍饲饲养是目前养猪业中广泛采用的一种工艺模式,虽然它具有占地少、便于饲养管理、可实现一定程度的环境调控等优点,但由于舍饲条件下环境相对贫瘠,猪只的福利较难得到保障。该文重点讨论了国内外生猪舍饲条件下相关因素与猪只福利之间的关系,分析了舍饲环境下因福利问题而带来的一些不利影响。根据目前国内外现有的研究结果,指出我国应加强舍饲条件下的猪只福利问题研究的必要性,并就改善舍饲条件下猪只福利的措施及相关配套设备进行了论述,为我国养猪业持续健康的发展、增强我国养猪业的国际竞争力提供了一些思路。     

Maturity indices for composted dairy and pig manures

Bulking agents and bedding materials used on farms for composting manures affect the time required for composts to mature. The effects of these materials on guidelines for the use of composted manures in potting mixes are not fully known. Several chemical and biological compost characteristics were mentioned and a cucumber plant growth greenhouse bioassay was performed on samples removed from windrows during composting of: (i) dairy manure amended with wheat straw; (ii) dairy manure amended with sawdust (mostly Quercus spp.); and (iii) pig manure amended with sawdust and shredded wood (mostly Quercus spp.). Dry weights of cucumber seedlings grown in fertilized and unfertilized potting mixes amended with composts (30%, v/v) having stability values of <1 mg CO₂-C g^-1 dw d⁻¹, did not differ significantly from those in a control peat mix. Only the most mature dairy manure-wheat straw compost samples consistently established sufficient N concentrations in cucumber shoots in unfertilized treatments. For the dairy manure-wheat straw compost, all possible subset regression analyses of compost characteristics versus cucumber plant dry weight revealed that any of several compost characteristics (electrical conductivity-EC, compost age, total N, organic C, C-to-N ratio, ash content, CO₂ respirometry, Solvita CO₂ index and the Solvita^® Compost Maturity Index) predicted growth of cucumber in the unfertilized treatments, and thus maturity. In contrast, at least two characteristics of the dairy manure-sawdust compost were required to predict growth of cucumber in the unfertilized treatments. Effective combinations were EC with compost age and the Solvita^® maturity index with total N. Even five compost characteristics did not satisfactorily predict growth of cucumber in the non-fertilized pig manure-wood compost. Nutrient analysis of cucumber shoots indicated N availability was the principal factor limiting growth in potting mixes amended with the dairy manure-sawdust compost, and even more so in the pig manure-wood compost even though the compost had been stabilized to a high degree (<1 mg CO₂-C g⁻¹ dw d⁻¹). Maturity of the composted manures, which implies a positive initial plant growth response of plants grown without fertilization, could not be predicted by compost characteristics alone unless the bulking agent or bedding type used for the production of the composts was also considered.     

舍饲条件下猪只福利的实现与设施配套

舍饲饲养是目前养猪业中广泛采用的一种工艺模式，虽然它具有占地少、便于饲养管理、可实现一定程度的环境调控等优点，但由于舍饲条件下环境相对贫瘠，猪只的福利较难得到保障。该文重点讨论了国内外生猪舍饲条件下相关因素与猪只福利之间的关系，分析了舍饲环境下因福利问题而带来的一些不利影响。根据目前国内外现有的研究结果，指出我国应加强舍饲条件下的猪只福利问题研究的必要性，并就改善舍饲条件下猪只福利的措施及相关配套设备进行了论述，为我国养猪业持续健康的发展、增强我国养猪业的国际竞争力提供了一些思路。     

Nitrogen losses following application of pig slurry to arable land

Abstract. Emissions of ammonia (NH₃) and nitrous oxide (N₂O), and nitrate (NO^-₃) leaching were measured in two field experiments following application of pig slurry at rates corresponding to 83–96 kg NH₄-N ha^-1 before sowing. In spring and in autumn 1994, slurry was applied by four methods: trenching (T), shallow injection (S), band spreading immediately followed by harrowing (B/H) and band spreading (B). NH₃ emission measurements were made during the first week after application in both experiments. In the spring experiment N₂O emissions and NO^-₃ leaching were measured during 6 and 52 weeks after spreading respectively, and during 11 and 33 weeks after spreading in the autumn experiment. In spring, the increased N₂O emissions (i.e. control subtracted) ranged from 0.27% (T) to 0.45% (B/H), and in the autumn study from 0.92% (T) to 1.14% (B/H), of applied NH₄-N, although showing no statistically significant differences. In order to validate the chamber measurements, a ‘megachamber’(21 m²) was used together with an infrared spectrometer. The emissions agreed well for (B/H), while (B) resulted in lower emissions compared with the smaller chambers. Emissions of NH₃ were about one order of magnitude higher. In spring, (B) gave the highest emission, reaching 19.5% of applied NH₄-N, whereas (S), and (B/H) gave the lowest emissions, reaching 1.2 and 3.5% of applied NH₄-N, respectively. NH₃ emissions in autumn were 15–20% lower compared with spring. In spring the increased nitrate leaching ranged from 10.1 (T) to 24.9 kg ha^-1 (B/H) and from 29.5 (B) to 37.8 kg ha^-1 (T) in the autumn experiment, showing no statistically significant differences. Estimations of indirect N₂O emissions due to ammonia deposition and nitrate leaching, suggested that the N₂O contribution from NH₃ deposition was relatively small, while the indirect N₂O emissions from NO^-₃ leaching were of the same order of magnitude or higher than the direct N₂O emissions.     

Microbial activity in pig slurry-amended soils under semiarid conditions

Soil amendment with manures from intensive animal industries is nowadays a common practice that may favorably or adversely affect several soil properties, including soil microbial activity. In this work, the effect of consecutive annual additions of pig slurry (PS) at rates of 30, 60, 90, 120 and 150 m³ ha⁻¹ y⁻¹ over a 4-year period on soil chemical properties and microbial activity was investigated and compared to that of an inorganic fertilization and a control (without amendment). Field plot experiment conducted under a continuous barley monoculture and semiarid conditions were used. Eight months after the fourth yearly PS and mineral fertilizer application (i.e. soon after the fourth barley harvest), surface soil samples (Ap horizon, 0-15 cm depth) from control and amended soils were collected and analysed for pH, electrical conductivity (EC), contents of total organic C, total N, available P and K, microbial biomass C, basal respiration and different enzymatic activities. The control soil had a slightly acidic pH (6.0), a small EC (0.07 dS m⁻¹), adequate levels of total N (1.2 g kg⁻¹) and available K (483 mg kg⁻¹) for barley growth, and small contents of total organic C (13.2 g kg⁻¹) and available P (52 mg kg⁻¹). With respect to the control and mineral fertilized soils, the PS-amended soils had greater pH values (around neutrality or slightly alkaline), electrical conductivities (still low) and contents of available P and K, and slightly larger total N contents. A significant decrease of total organic C was observed in soils amended at high slurry rate (12.3 g kg⁻¹). Compared with the control and mineral treatments, which produced almost similar results, the PS-amended soils were characterized by a higher microbial biomass C content (from 311 to 442 g kg⁻¹), microbial biomass C/total organic C ratio (from 2.3 to 3.6%) and dehydrogenase (from 35 to 173 μg INTF g⁻¹), catalase (from 5 to 24 μmol O₂ g⁻¹ min⁻¹), BAA-protease (from 0.7 to 1.9 μmol  g⁻¹ h⁻¹) and β-glucosidase (from 117 to 269 μmol PNP g⁻¹ h⁻¹) activities, similar basal respirations (from 48 to 77 μg C-CO₂ g⁻¹ d⁻¹) and urease activities (from 1.5 to 2.2 μmol  g⁻¹ h⁻¹), and smaller metabolic quotients (from 6.4 to 7.7 ng C-CO₂ μg⁻¹ biomass C h⁻¹) and phosphatese activities (from 374 to 159 μmol PNP g⁻¹ h⁻¹). For example, statistical analysis of experimental data showed that, with the exception of metabolic quotient and total organic C content, these effects generally increased with increasing cumulative amount of PS. In conclusion, cumulative PS application to soil over time under semiarid conditions may produce not only beneficial effects but also adverse effects on soil properties, such us the partial mineralization of soil organic C through extended microbial oxidation. Thus, PS should not be considered as a mature organic amendment and should be treated appropriately before it is applied to soil, so as to enhance its potential as a soil organic fertilizer.     

猪宰后肌肉中钙激活蛋白酶活性变化的研究

该文研究杂种野猪和本地白猪屠宰后，肌肉中钙激活蛋白酶系的活性变化，同时，对肌肉的全蛋白提取液作SDS-PAGE分析。结果表明：杂种野猪肉的钙激活蛋白酶活性较高，但表现酶活性的自溶率变化在早期较慢；电泳图谱可看到，杂种野猪蛋白降解较慢，在宰后冷藏40 h的电泳图可看到小分子物质降解较少。     

The effect of pig slurry on exchangeable potassium in calcareous soils

Pig slurry in quantities of 200, 400, 500, 600, 800, and 1000 m³ ha^-1 year^-1 was added to two calcareous soils, which differed in concentration and type of clay minerals. The soils were cultivated with two successive crops of pepper and tomatoes grown in containers. A control was given no slurry. The soils were analysed after harvesting for exchangeable K⁺. Differences in exchangeable K⁺ were studied in relation to the concentration and type of clay minerals. The soil with the higher clay content and of the illite type retained K in the exchangeable form to a much greater extent than the soil with the low clay content. In the soil with the lower clay content, of the interstratified illitic — montmorillonite type, very little K was incorporated into the exchange complex. The exchange capacity being low, the amount of K added had little effect on the level of exchangeable K⁺.     

Dust in pig buildings

It is well documented in the international scientific literature that airborne dust in pig houses can cause serious health problems for humans as well as for animals. Extensive research has been carried out in different countries during the last few decades to improve the scientific understanding of air quality issues related to intensive animal production. Research and review papers were presented at the international symposium on Dust Control in Animal Production Facilities, held in Denmark in 1999. Different techniques have been used in order to reduce dust burdens in pig confinement buildings, but up to date only the procedure of spraying oil or a mixture of oil and water has contributed to reducing the indoor dust concentrations significantly. This article summarizes the current level of understanding of dust issues in intensive animal production buildings, mainly on the basis of papers presented at the above-mentioned symposium.