基于时间序列分析的卡尔曼滤波组合导航算法

GPS广泛用于农业机械导航研究中,其定位误差信号一般存在明显的自相关性,不能满足组合导航中常用的卡尔曼滤波算法观测噪声为高斯白噪声的要求。为此,建立了GPS定位误差AR模型,结合卡尔曼估计结果来预测和修正GPS定位误差,再将修正后的GPS定位信息应用于组合导航中的卡尔曼滤波过程。试验结果表明,无论GPS接收机是在静止还是在运动条件下,处理后的定位误差信号自相关性都明显降低,近似为白噪声;目标路径直线时的最大跟踪误差约为0.15 m,为曲线时,最大跟踪误差约为0.3 m。该方法为低精度GPS应用于农业机械导航提供了可行途径。     

基于Gray-EKF算法的智能农业车辆同时定位与地图创建

为了提高智能农业车辆在未知环境中同时定位与地图创建精度,将灰色预测理论和扩展卡尔曼滤波融合,提出了基于灰色扩展卡尔曼滤波的同时定位与地图创建算法。算法在传统的扩展卡尔曼滤波基础上,通过改进的滑窗灰色预测理论建立传感器的GM(1,1)观测预测模型,进而完成新息的计算。为了提高观测精度和抗干扰能力,系统使用了三目摄像机作为观测传感器,并提出了一种简易的权值标定算法。试验表明:精度权值标定后的三目摄像机具有较高的测量精度,16组测量数据中有12组的测量误差小于1%,并能减小由于干扰造成的误差。在30个人工路标的停车场环境中,车辆对路标x和y方向的观测误差均值为0.074和0.073m,自身定位误差为0.140m,误差均方差为0.048。在60个人工路标的停车场环境中,车辆对路标x和y方向的观测误差均值为0.061和0.068m,自身定位误差为0.109m,误差均方差为0.038。在60个人工路标的旱地环境中,车辆对路标x和y方向的观测误差均值为0.079和0.077m,自身定位误差为0.122m,误差均方差为0.049。研究认为,与传统的EKF SLAM算法相比,Gray-EKF SLAM算法具有更高的精度。     

构建VED-SegNet分割模型提取鱼类表型比例

为提高鱼类表型分割精度和准确度,实现鱼类表型智能监测,该研究基于深度学习算法构建了VED-SegNet模型用于鱼类表型分割和测量。该模型将cross stage partial network和GSConv结合作为编码器（VoV-GSCSP）,保持足够精度的同时降低网络结构复杂性。另一方面,该模型采用EMA(efficient multi-scale attention module with cross-spatial learning)建立强化结构,加强编码器和解码器之间的信息传递,提高模型精度,并实现了8个表型类别的输出。采用自建的鱼类表型分割数据集对VED-SegNet模型进行了测试,测量结果中鱼类各表型比例与实际测量值相接近,表型最大平均绝对和平均相对误差为0.39%、11.28%,能实现无接触式提取水产养殖中鱼类表型比例。对比其他常见语义分割模型,平均交并比mean intersection over union,mIoU和平均像素准确率mean pixel accuracy,m PA最高,分别到达了87.92%、92.83%。VED-SegNet模型在环境复杂、多鱼重叠的...     

基于LIO-SAM建图和激光视觉融合定位的温室自主行走系统

为解决传统导航方案在温室内无法应对光照变化大、作物行间距窄、接收GPS信号差等问题,该研究提出了基于即时定位与地图构建技术的激光视觉融合式自主导航算法。该系统利用三维激光雷达VLP-16(Velodyne LiDAR,VLP-16)和惯性测量单元获取温室环境信息,采用基于紧耦合的雷达惯导定位建图（tightly-coupled lidar inertial odometry via smoothing and mapping,LIO-SAM）算法构建导航地图,基于轮式里程计和视觉里程计采用扩展卡尔曼滤波器算法实现局部定位,融合激光点云配准算法和自适应蒙特卡洛定位算法实现全局定位。同时,在自主行走系统应用A*算法规划全局路径和动态窗口算法规划局部路径,从而实现自主导航。试验结果表明,LIO-SAM算法构建的温室导航地图最大相对误差、最大绝对误差和均方根误差分别为9.9%、0.081和0.063 m,在温室内改进后的定位算法横向偏差小于0.020 m,纵向偏差小于0.090 m;当自主行走系统以0.15、0.30和0.50 m/s的速度运行时,横向偏差、纵向偏差和航向偏角的平均值分别小于0...     

珠江三角洲河网区水体及鱼体内重金属含量分析与评价

采用原子吸收及原子荧光法,对珠江三角洲河网区水体及鱼体中重金属Cr、Cu、Pb、Fe、Zn、Mn、Ni、Cd、As和Hg含量进行了检测,并以国家地表水环境质量标准（Ⅲ类）及水产品中有毒有害物质限量进行了评价。结果表明,水体中Cr、Pb和Hg为主要污染元素,超标率分别为22.2%、11.1%和5.6%,水体中不同季节Pb和Cr含量变化基本特征为丰水期〉平水期〉枯水期,河网不同江段变化特征不同,这与重金属在水体中的存在形态及受水动力条件影响有关;鱼体中残留的重金属元素有Cr、Pb和As,超标率分别为36.7%、27.1%和6.3%,超标鱼类主要分布于肇庆-富湾及大塘-三水江段。不同鱼类重金属元素残留特征为底层鱼大于中上层鱼,肉食性鱼大于植食性鱼和杂食性鱼。Pb和Cr为珠江三角洲河网区水体及鱼体中重金属污染的主要元素。     

基于鱼类上溯行为的导鱼堰设计与效果评估

河道内鱼类上溯路径不唯一,聚集位置更是难以预测,对鱼类上溯行为进行有效引导有助于提升过鱼设施进口的效率。为此,该研究提出了一种导鱼堰的概念和设计方法,结合姚家坪水电站的过鱼设施,利用流场三维数值模拟、鱼类洄游（active fish migration,AFM）模型和实鱼试验对导鱼堰的效果进行评估。结果表明：姚家平水利枢纽工程的导鱼堰上下游水面落差为0.36～0.40 m,过堰水流流速可达1.5～2.8 m/s,形成阻鱼的屏障,并在导鱼堰下游侧形成了诱导鱼类向集鱼渠进鱼口游动的唯一低流速上溯通道,鱼类聚集点趋于唯一,验证了导鱼堰方案的合理性。利用鱼类洄游模型对导鱼堰的导鱼效果进行预测,结果表明,下游高水位和低水位2种工况下,90%以上的鱼类游动路径均表现出相似的规律,鱼类沿河道右岸和导鱼堰下游侧的低流速通道上溯,并最终聚集在集鱼渠的进鱼口处。放鱼试验中试验个体全部进入集鱼渠,结果进一步证实导鱼堰可以有效引导鱼类游动路线,并在集鱼渠进鱼口处形成唯一的聚集区。本文提出的导鱼堰丰富了生态水工建筑物的形式,可为过鱼设施进口的水力设计及下游河道的局部整治提供参考。     

基于蜂群优化多核支持向量机的淡水鱼种类识别

为了准确地进行淡水鱼种类自动识别,利用计算机视觉技术,提出了一种基于Krawtchouk矩、灰度共生矩阵、蜂群优化多核最小二乘支持向量机（least squares support vector machine,LS-SVM）的识别方法。首先获取淡水鱼样本的灰度图像,计算淡水鱼鱼体的长宽比、鱼头鱼尾的Krawtchouk矩不变量形状特征,求得鱼身的灰度共生矩阵纹理特征,将上述形状与纹理特征组合成高维特征向量,并输入到多核LS-SVM,通过人工蜂群（artificial bee colony,ABC）算法对多核LS-SVM中的待定参数进行寻优,ABC算法中的适应度函数为测试样本的识别精度;最后输出识别精度达到最高时的最优参数。利用该方法对鳊鱼、鳙鱼、鲫鱼、草鱼、青鱼5种淡水鱼进行了分类识别,对鳊鱼、鳙鱼、鲫鱼、草鱼4种鱼识别时,各类鱼的识别精度均达到95.83%以上,对鳊鱼、鳙鱼、鲫鱼、青鱼4种鱼识别时,各类鱼的识别精度均达到91.67%以上,对鳊鱼、鳙鱼、鲫鱼、草鱼和青鱼 5种鱼识别时,各类鱼的识别精度均达到83.33%以上;与近年来提出的淡水鱼识别方法、BP（back propagation）神经网络方法、单核LS-SVM方法相比,该方法的识别精度更高,从而可快速准确地识别淡水鱼的种类,提高水产养殖的自动化水平。     

基于计算机视觉技术的运动鱼检测算法

为了提高水中鱼的运动检测精度,该文在分析比较图像和视觉领域常用的运动目标检测算法的基础上,提出了3种适合于运动鱼检测的改进算法。试验结果表明,该文提出的算法在检测鱼的快速运动方面要优于传统算法,特别是帧间差分和运动平均背景建模结合方法较其他算法在平均检测时间和检测精度方面更具优势,在鱼类运动检测方面有较大的应用潜力。     

基于机器视觉的鲐鱼鱼体定向排列输送装置设计与试验

针对传统鱼体头尾及腹背定向输送由人工操作完成,劳动强度大、生产效率低等问题,该研究探索了利用机器视觉技术结合输送装置实现鱼体头尾及腹背定向排列输送的方法。该研究以鲐鱼(Scomber japonicus)为研究对象,在对鱼体形态特征及物理特性检测的基础上,设计了鱼体提升装置、鱼体分离输送装置、鱼体头尾及腹背定向输送装置、鱼体返回输送装置、定向控制系统等部件组成鲐鱼鱼体定向排列输送装置;构建YOLOv5s目标检测模型对鲐鱼的头尾朝向和腹背朝向进行识别,并根据识别结果控制鱼体定向排列输送完成鲐鱼的头尾和腹背定向作业,检测模型在测试集上的精确率为99.76%,召回率为99.59%,平均检测精度值为99.5%;试制了鱼体定向排列输送装置样机,以单条鱼体提升成功率为评价指标,对不同输送速度下鱼体提升装置的输送效果进行试验;同时以鱼体的头尾定向成功率和腹背定向成功率作为评价指标,以鱼体提升装置输送速度、鱼体分离输送装置输送速度、鱼体头尾及腹背定向输送装置输送速度为试验因素,对鱼体定向排列输送装置的定向输送效果进行试验。试验结果表明：鱼体提升装置在不同输送速度下,都能有效实现鱼体单条分离并向上提升,且不存在鱼体重叠向上输送的情况;当鱼体提升装置输送速度为0.05 m/s、鱼体分离输送装置输送速度为0.45 m/s、鱼体头尾及腹背定向输送装置输送速度为0.60 m/s时,鲐鱼鱼体的头尾定向成功率平均为97.2%,腹背定向成功率平均为95.6%,鱼体定向输送速度可达15条/min。研究结果可为其他淡水鱼鱼体定向排列输送装置的研制提供参考。     

鱼感应流速对体长的响应及在过鱼设施流速设计中的应用

该研究测试并分析了澜沧江5种鱼类(体长bl的范围0.095~0.342 m)的感应流速。测试方法:首先测量鱼的长度,然后将其放于鱼类行为学装置内的游泳区中,并使其适应2 h。然后让鱼的头部朝着游泳区后端,以递增流速法调整装置内水流速度。当鱼类调转方向运动时,记录此时装置内水流速度,该水流速度即为鱼类感应流速。研究结果:1)在水温为11.1~22.6℃条件下,鱼类感应流速范围为(0.40~1.18)bl/s,即(0.07~0.23)m/s;2)鱼类相对感应流速和体长总体呈负相关关系,鱼类绝对感应流速和体长总体呈正相关关系,该相关关系对于较难捕获测试样本或仅可捕获极个别样本的鱼类游泳特性的定性预测研究具有一定指导作用;3)拟合了趋流率与水流速度的相关关系,可为寻求满足鱼类游泳需求和工程流速设计上的平衡提供参考,若仅考虑过鱼设施运行效率,则澜沧江上游某水电工程过鱼设施内最低设计流速至少为0.2 m/s。该研究结果可为鱼类行为学提供基础资料并为过鱼设施流速设计提供参考。     

Watershed risk analysis model for TVA's Holston River basin

The Electric Power Research Institute has launched a research project to develop a conceptual risk analysis framework for watershed management of point and nonpoint source pollution. The research leads to the design of an engineering model to 1) process and translate water quality data (coliform, HOD, DO, suspended solids, temperature, sediment, etc.) into decision variables (suitability for water contact sports and swimming, fish spawning fish survival, human consumption of fish, and freedom from algal nuisance, etc.) and 2) predict water quality improvements from proposed management alternatives. Actual development of the model is being carried out with the Tennessee Valley Authority (TVA) for the Holston River watershed.The effort includes model construction by importation of GIS map files, stringing together existing watershed and reservoir models, calibration of the model, and selection of decision variables and water quality check points. The model calculates hydrology, waste load, water quality and suitability of fish habitats at headwaters. The base case results and improvements after best management alternatives will be compared to the data observed by TVA's River Action Team. The final product will be a user friendly tool that stakeholders can use to find a cost effective method of improving water quality, including market-based pollution trading.     

Influence of connectivity, habitat quality and invasive species on egg and larval distributions and local abundance of crucian carp in Japanese agricultural landscapes

The quality of habitat at and around a spawning site, and the availability of movement between spawning and nonspawning habitats are likely to be important determinants for population persistence in a degraded and fragmented landscape. We assessed the influence of habitat connectivity, habitat quality and invasive species for distributions and local abundance of eggs and larvae of crucian carp (Crassius auratus complex, which is listed as “data deficient” on the Japanese Red List) in agricultural landscapes surrounding Lake Mikata, Japan, where drainage ditches and paddy fields are extensively utilised for spawning (lake or river shores are also used). We investigated the presence and abundance of eggs and larvae of crucian carp and habitat components at 146 sites across a range of presumed spawning habitats. Egg presence was affected strongly by connectivity to the lake (watercourse distance from the lake), and egg abundance was significantly influenced by both connectivity and habitat quality. In contrast, larval presence was primarily related to habitat quality. Larval abundance was influenced by connectivity and habitat quality, but the effect of connectivity was relatively low. Furthermore, larval abundance was negatively related to the presence of the invasive species red swamp crayfish (Procambarus craki) and bullfrog (Rana catesbeiana). Our findings indicate that connectivity, habitat quality and the presences of invasive species are crucial in determining suitable spawning and nursery habitats, but their relative importance may vary depending on egg and larval life stages. We suggest that restoring connectivity, improving habitat quality and removal of invasive species could be effective conservation strategies for the declining populations of crucian carp in agricultural landscapes.     

人工坝体对长江上游鱼类栖息地流域水动力学特性的影响

长江上游鱼类栖息地的水动力学研究对于栖息地生态治理与恢复、评估水生物的生境条件有重要科学意义。该研究以长江上游四川省泸州市江段人工修复鱼类栖息地为研究对象,根据现场施工图、水下地形测绘图和航拍图,等比例建立了三维流场模型。基于CFD数值模拟结合现场试验勘测,研究了6种不同流速下的栖息地水动力学特性,分析了多个人工坝体对该江段水流环境的影响机制、湍流/涡结构的演化与发展规律,利用了水流均匀性指数、傅汝德数、局部断面流速等水力指标对修复后的栖息地水环境进行评估,并与现场地形和流速测绘、鱼群调查等试验数据对比分析。研究表明:在该江段试验区形成多个大漩涡区,涡流形态基本不随来流速度的变化而改变,逆时针漩涡区的上升流现象明显,且流速稍高于顺时针漩涡区;试验区的水流速度偏低,流动均匀性差,流态紊乱复杂,而对照区的流速相对较快,流动均匀性好,水流方向变化不明显;随来流速度增加,傅汝德数不完全呈线性增加趋势,不同水深截面上的水流均匀性指数的变化趋势相近;现场鱼群调查相同测绘距离内,对照区域为0.714尾/1000呯,试验区域为3.157尾/1000呯,说明流动不均匀的低速涡流区有利于聚集鱼群,经修复后的栖息地能够为水生物提供良好的栖息、繁衍、庇护等生存条件;当来流速度为2 m/s时,试验区域两条测试路径的平均流速分别为0.53和0.79 m/s,在同一鱼群探测试验测绘路径上,数值模拟结果与现场试验数据相差约10%,验证了数值计算方法的可靠性。研究结果对于大水面渔业生态工程的水动力学研究有重要的参考意义。     

The loch fleet and other catchment liming programs

The circumstances under which catchment liming is an advisable economic alternative to direct water liming are described, as are research and operational experiences of catchment liming, drawing mainly on the Loch Fleet Project results and other European data. The arguments for and against catchment liming as a means of restoring acidified aquatic habitat are presented. In general, it is concluded that there are conditions, such as lakes with short retention times and/or where stream spawning fish species are involved, where catchment liming is a sensible option.     

基于自适应有限冲激响应-卡尔曼滤波算法的GPS/INS导航

导航定位系统一般采用卡尔曼滤波算法提高定位精度。传统卡尔曼滤波算法的性能很大程度上依赖观测噪声的先验统计信息,不精确的统计特性将会降低定位精度。针对此问题,该文提出一种基于FIR(finite impulse response)预测模型的卡尔曼滤波算法。将FIR预测模型与卡尔曼滤波结合,FIR预测模型的系数可以通过求解一个凸二次规划问题得到。该凸二次规划以目标的多项式运动规律为约束条件,以最小白噪声增益为目标函数,具有闭式解。仿真试验和实测结果均表明,在相同的参数设置条件下,基于FIR预测模型的卡尔曼滤波算法比传统的卡尔曼滤波算法具有更高的估计精度,仿真结果表明定位精度提高29.54%,实测结果表明X方向定位精度提高21.71%,Y方向定位精度提高22.62%。该算法可应用于GPS接收信号的降噪处理,提高目标状态的定位精度。     

Captive breeding promotes aggression in an endangered Mexican fish

We compared the behaviour of wild and captive-bred butterfly splitfins (Ameca splendens), an endangered freshwater fish, to investigate whether captive breeding results in the behavioural divergence of wild and captive individuals. In a first experiment, we examined whether the captive environment allows for the similar expression of behaviours observed in wild fish. The foraging, courtship and aggressive behaviour of fish in their natural habitat (in Mexico) was compared with that of their counterparts that have been bred at London Zoo, UK, for 40 years. These in situ observations revealed that wild fish were preoccupied with searching for food whereas captive fish engaged more in aggressive interactions. In a subsequent laboratory experiment we compared the behaviour of wild-caught and captive-bred fish under standard conditions in two novel habitats: structured (enriched) and unstructured (bare) aquaria. Overall, captive-bred butterfly splitfins displayed higher levels of aggression than wild-caught fish. The relationship between aggression and habitat structure was influenced by density; captive-bred males were more aggressive when observed in structured habitats than unstructured ones, but only when they were stocked at a high density. We also found an effect of tank structure on foraging behaviour, with individuals spending more time foraging in unstructured tanks than structured tanks. There was no effect of captive breeding or habitat structure on courtship behaviour.Our findings suggest that captive environments can promote the development of aggressive behaviour which may affect the suitability of captive-bred fishes for reintroduction into the wild.     

荔枝采摘机器人双目视觉的动态定位误差分析

扰动引起的随机误差成为采摘机器人视觉定位的难题。为了探索荔枝采摘机器人视觉定位误差,首先用双目视觉系统和模拟扰动的震动平台对荔枝结果母枝采摘点的三维坐标进行定位试验,检测其实际位置,获得误差数据;然后,提出了一种动态定位误差分析方法,根据误差变化规律将动态定位误差划分为系统误差和随机误差;最后,用统计方法对2类误差分别进行定量分析和评价。结果表明,定位距离为600~1 000 mm时,系统误差与动态定位误差的变化趋势基本一致,视觉深度方向、水平方向最大动态定位误差分别为58.8和17.3 mm。系统误差置信区间较窄,视觉深度方向系统误差与定位距离呈较强的线性相关性,水平方向则表现为非线性。扰动下的随机定位误差服从正态分布,视觉深度方向、水平方向间的随机误差相关性较弱。视觉深度方向受扰动的影响较大,随机误差远大于水平方向,且不确定度较高。研究结果为荔枝采摘机器人视觉定位系统校准和动态定位方案设计提供依据,为机构容错纠错提供理论依据和实践指导。     

Factors Affecting Brain Cholinesterase Activity in Bluegill (Lepomis Macrochirus)

Organophosphorus insecticides (OPs) are widely used to control crop pests, and can enter aquatic systems draining agriculturalwatersheds. Because OPs are highly toxic to aquatic organisms,monitoring is important, yet difficult due to the low persistenceof many OPs. Cholinesterase (ChE) inhibition can be used to monitor organisms' exposure to OPs and carbamates; however, highvariability in ChE activity may interfere with the ability of theassay to detect OP exposure. Several potential sources of variation in ChE activity, including water temperature, storage of samples, euthanasia method, and fish sex and size, were investigated in bluegill (Lepomis macrochirus). ChE activity did not vary significantly with water temperature in the range 20–31 °C, or following short-term storage ofeuthanized bluegill up to 8 h at 4 and 19 °C or long-term storage of brain homogenates up to 389 d at –198 °C. There was no difference in ChE activity by euthanasia method (spinal cord severing or anesthesia with Finquel®), nor by sex of fish. Mean ChE activity variedinversely with fish total length. Of the variables investigated in this study, only size needs to be restricted in fish collectedfor monitoring of OPs.     

基于视觉伺服的联合收割机群协同导航从机定位方法

针对收割机群协同导航从机实际定位需求,在分析主机与从机间相对位置关系的基础上,提出了一种采用视觉系统求解靶标投影矩形区域形心坐标,并以此作为导引信息,引导相对位姿测量装置测量从机相对主机位姿,实现从机精确定位的方法。重点对从机定位方法中需解决的运动靶标跟踪和对靶控制2个关键问题进行了研究,提出了一种基于双重波门的运动靶标跟踪方法和一种基于变尺度变论域模糊控制的对靶控制方法。试验结果表明,该文所提定位方法基本不受收割机前进速度的影响,不同速度最大偏差的标准差为0.003 5 m,平均偏差的标准差为0.002 3 m。     

Deforesting the riverscape: the effects of wood on fish diversity in a Venezuelan piedmont stream

While deforestation of tropical ecosystems has been shown to have significant impacts on terrestrial habitats, its effects on aquatic habitats are poorly studied. Deforestation dramatically reduces the input of woody debris to streams, and given the importance of large woody debris to fish communities in temperate streams, this might be one mechanism by which logging could affect aquatic ecosystems in the tropics. To examine the effects of large woody debris on the diverse fish assemblage of a tropical stream, we surveyed pools with and without wood at Rio Las Marias, Venezuela. Pools containing wood contained greater numbers of individuals and more species of fish than pools without wood, and the two types of pools differed in their composition. To test whether these results were due to the presence of woody debris, we conducted an experimental wood addition. Pools to which wood was added showed marked increases in both fish abundance and species richness relative to wood-free pools, and the composition of the fish assemblage in experimental pools approached that of pools with naturally occurring woody debris. These results demonstrate that large woody debris plays a major role in structuring fish communities in tropical streams. As a consequence, logging practices that reduce the input of woody debris to tropical streams or directly remove wood from streams could have serious impacts on aquatic habitats, affecting both the diverse fish communities and local economies dependent on stream fisheries.