Alpha-shape算法构建枣树点云三维模型

为了实现枣树智能化修剪作业，该研究提出了基于点云配准的自然光照环境下的果树三维重构方法，并针对传统最近点迭代（Iterative Closest Point，ICP）算法对待配准点云的空间位置要求苛刻的问题，提出了改进的点云配准算法。首先，使用彩色深度（RGB-D）相机采集不同角度下的枣树彩色和深度图像，并通过信息融合实现相应角度下的点云获取。其次，对点云进行背景去除和滤波处理，基于直方图设定分割阈值，提取单株枣树点云，并将放置在树根附近的标靶球作为标记，使用人工标记法进行两站点云初配准。最后，在初配准基础上计算点云的曲面法向量和曲率，由曲率相近的点构成配对点对，使用k维树最近点迭代（k dimensional-tree-Iterative Closest Point，kd-tree-ICP）算法完成精配准，对点云使用Alpha-shape算法面片化，实现表面重构。利用上述方法对多棵枣树进行全局配准并完整重构果树模型。试验结果表明，通过引入初配准，有效提高了点云配准的准确性和稳定性，配准误差均控制在1.0 cm以内，平均配准误差为0.78 cm；重构模型真实感较强，在外观上更加接近真实树，枝干相对误差控制在7%以内。该研究重构模型精度较高，可为枣树智能修剪提供可视化研究基础和技术支持。     

Deep neural network algorithm for estimating maize biomass based on simulated Sentinel 2A vegetation indices and leaf area index

Accurate estimation of biomass is necessary for evaluating crop growth and predicting crop yield.Biomass is also a key trait in increasing grain yield by crop breeding.The aims of this study were(i)to identify the best vegetation indices for estimating maize biomass,(ii)to investigate the relationship between biomass and leaf area index(LAI)at several growth stages,and(iii)to evaluate a biomass model using measured vegetation indices or simulated vegetation indices of Sentinel 2A and LAI using a deep neural network(DNN)algorithm.The results showed that biomass was associated with all vegetation indices.The three-band water index(TBWI)was the best vegetation index for estimating biomass and the corresponding R2,RMSE,and RRMSE were 0.76,2.84 t ha−1,and 38.22%respectively.LAI was highly correlated with biomass(R2=0.89,RMSE=2.27 t ha−1,and RRMSE=30.55%).Estimated biomass based on 15 hyperspectral vegetation indices was in a high agreement with measured biomass using the DNN algorithm(R2=0.83,RMSE=1.96 t ha−1,and RRMSE=26.43%).Biomass estimation accuracy was further increased when LAI was combined with the 15 vegetation indices(R2=0.91,RMSE=1.49 t ha−1,and RRMSE=20.05%).Relationships between the hyperspectral vegetation indices and biomass differed from relationships between simulated Sentinel 2A vegetation indices and biomass.Biomass estimation from the hyperspectral vegetation indices was more accurate than that from the simulated Sentinel 2A vegetation indices(R2=0.87,RMSE=1.84 t ha−1,and RRMSE=24.76%).The DNN algorithm was effective in improving the estimation accuracy of biomass.It provides a guideline for estimating biomass of maize using remote sensing technology and the DNN algorithm in this region.     

基于特征金字塔注意力与深度卷积网络的多目标生猪检测

在生猪饲养环境中,猪只黏连、杂物遮挡等给生猪个体多目标检测带来很大困难。该研究以猪圈群养生猪为研究对象,以视频帧为数据源,提出一种适用于生猪形体检测的特征金字塔注意力(FeaturePyramidAttention,FPA)与Tiny-YOLO相结合的模型FPA-Tiny-YOLO。该模型将注意力信息融入到特征提取过程,在不大幅增加计算量的前提下即可提升特征提取能力、提高检测精度。对8栏日龄20～105 d的45头生猪视频截取图像进行图像处理,获得标注图片4 102张,构建了4种深度FPA模块分别加入YOLOV3与Tiny-YOLO模型中。试验表明,深度为3的FPA模块(即FPA-3)的Tiny-YOLO模型在测试集上对群养生猪多目标检测的召回率Recall、F1与平均检测精度m AP指标值最佳,分别达到86.09%、91.47%和85.85%,比未引入FPA模块的Tiny-YOLO模型均有不同程度的提高。选用不同的IOU(Intersection Over Union)和score阈值超参数值对模型预测结果均有不同程度影响;将测试集图像按照是否黏连与遮挡划分4种场景来探究该模型的鲁棒性。试验表明,加入FPA-3模块后Tiny-YOLO的Recall、F1与m AP比Tiny-YOLO分别提升6.73、4.34和7.33个百分点,说明特征金字塔注意力信息有利于精确、有效地对不同场景群养生猪进行多目标检测。研究结果可为后续开展生猪身份识别和行为分析移动端应用提供参考。     

Land surface temperature retrieval for arid regions based on Landsat-8 TIRS data:a case study in Shihezi,Northwest China

Scientific interest in geophysical information about land surface temperature （LST） is ever increasing, as such information provides a base for a large number of applications, including environmental and agricultural monitoring. Therefore, the research of LST retrieval has become a hot topic. Recent availability of Landsat-8 satel- lite imagery provides a new data source for LST retrieval. Hence, exploring an adaptive method with reliable ac- curacy seems to be essential. In this study, basing on features of Landsat-8 TIRS thermal infrared channels, we re-calculated parameters in the atmospheric transmittance empirical models of the existing split-window algorithm, and estimated the ground emissivity with the help of the land cover classification map of the study area. Further- more, a split-window algorithm was rebuilt by virtual of the estimation model of the updated atmospheric transmit- tance and the ground emissivity, and then a remote sensing retrieval for the LST of Shihezi city in Xinjiang Uygur autonomous region of Northwest China was conducted on the basis of this modified algorithm. Finally, precision validation of the new model was implemented by using the MODIS LST products. The results showed that the LST retrieval from Landsat-8 TIRS data based on our algorithm has a higher credibility, and the retrieved LST is more consistent with the MODIS LST products. This indicated that the modified algorithm is suitable for retrieving LST with competitive accuracy. With higher resolutions, Landsat-8 TIRS data may provide more accurate observation for LST retrieval.     

基于无人机高光谱遥感的柑橘患病植株分类与特征波段提取

目的 结合传统与现代农业病虫害监测的优缺点，探索通过无人机高光谱遥感技术检测出患病的柑橘植株、通过人工田间调查方式判断其患病种类及患病程度的病虫害监测方法。方法 使用无人机获取原始高光谱图像，经过光谱预处理和特征工程后，采用连续投影算法提取对柑橘患病植株分类贡献值最大的特征波长组合，基于全波段使用BP神经网络和XgBoost算法、基于特征波段使用逻辑回归和支持向量机算法，建立分类模型。结果 基于全波段的BP神经网络和XgBoost算法的ROC曲线下面积(Area under curve，AUC)分别为0.883 0和0.912 0，分类准确率均超过95%；提取出698和762 nm的特征波长组合，基于特征波长使用逻辑回归和支持向量机算法建立的分类模型召回率分别达到了93.00%和96.00%。结论 基于特征波长建模在患病样本分类中表现出很高的准确率，证明了特征波长组合的有效性。本研究结果可为柑橘种植园的病虫害监测提供一定的数据和理论支撑。     

基于记忆模拟退火和A*算法的农业机器人 遍历路径规划

【目的】解决农业机器人大田作业时遍历路径规划的问题。【方法】提出一种记忆模拟退火与A*算法相结合的遍历算法。首先通过记忆模拟退火算法搜索出任务最优目标点行走顺序,然后使用A*算法进行跨区域衔接路径规划。【结果】仿真试验结果表明,该算法规划的遍历路径曼哈顿距离比传统模拟退火算法减少了9.4%,遍历路径覆盖率能达到100%,重复率控制为4.2%。【结论】记忆模拟退火通过为传统模拟退火算法增加记忆器,增强了跳出局部最优陷阱的能力,提高了算法所得解的质量。该研究结果可为农业机器人遍历路径规划提供理论基础。     

改进RBF神经网络在我国大豆价格预测中的应用研究

我国大豆价格受国内外多种因素共同影响,具有非线性、随机性和高噪音等特点,采用传统数学模型进行预测,不仅分析难度大,预测误差也很大。RBF神经网络以其优良的逼近性能而被广泛应用于非线性时间序列预测之中。本文提出一种基于遗传算法优化RBF神经网络的我国大豆价格预测模型,该模型为多维输入单维输出的多变量预测模型,模型的初始输入由大豆价格的历史数据和相关影响因素数据组成。采用遗传算法对RBF神经网络输入层节点数、基函数中心、扩展常数和输出层权值进行优化,模型可以从初始输入变量中自主选择最合适的输入变量组合作为模型的输入。采用2009-2014年的大豆价格数据进行预测研究,用2009-2013年的数据作为训练集,2014年的数据作为测试集,改进RBF神经网络通过自主识别和选取中国大豆进口量、中国消费者信心指数和进口大豆到港分销价格3个因素作为相关影响因素的输入。结果表明:模型预测精度较高、泛化能力较强,能够很好捕捉大豆价格变化规律,可为大豆市场价格的准确预报提供参考借鉴。     

基于随机森林算法和气象因子的砀山酥梨始花期预报

准确预报始花期是制定砀山酥梨花期管理措施和赏花活动方案的重要基础。该文利用1983-2018年砀山酥梨始花期的定位观测物候数据和平行观测的气象资料，采用线性趋势法，揭示始花期演变趋势；采用相关分析，筛选影响始花期的关键气象因子，依据不同预报日期构成特征变量集；采用随机森林算法（Random Forest, RF），自3月11日开始预报到3月25日终止预报，每日训练1个预报模型。结果表明，1）1983－2018年始花期呈极显著提早发生趋势，每10a约提前2.750 d（P＜0.001）。2）16个逐日气象预报模型中，共计有200个气象因子与始花期早迟密切相关，相关系数在0.469～0.789之间；各气象预报模型的训练集与测试集的平均正确率（Nd）分别为92.9%和75.5%、平均均方根误差（RMSE）分别为1.693～2.870和2.240～7.237、平均决定系数（R2）分别为0.891和0.701。3）2019年试验预报中，提前15日准确预报出当年始花期。该文研究表明RF在梨树始花期逐日气象预报中有一定业务应用潜力，预报准确率基本满足气象服务需求。     

基于模糊神经网络的智能温室环境控制方案

针对农业温室环境的精确建模和控制问题,提出了一种基于模糊神经网络的智能控制方案。首先,在考虑室内外环境因素下,构建一个有效的温室环境数学模型,获得通风量、喷雾量和加热量的微分表达式;然后,利用一种自适应模糊神经推理系统(ANFIS),以温度和湿度差作为输入,通过神经网络自学习和模糊推理获得控制输出;最后,通过遗传算法优化控制器的输出比例因子,提高控制响应速度和稳定性。实验结果表明:该方案能够快速且稳定地追踪环境设置值,具有很好的控制效果。     

New Approaches to Modeling Methyl Jasmonate Effects on Pomegranate Quality during Postharvest Storage

Adaptive neuro fuzzy inference system (ANFIS) and genetic algorithm–artificial neural network (GA-ANN) models were used to predict the effect of methyl jasmonate (at three levels 0, 0.01, and 0.1 mM) and storage time (0, 14, 28, 42, 56, 70, and 84 days) on quality parameters and physiological changes of pomegranate fruits during storage. Methyl jasmonate reduced chilling injury and improved quality characteristics of pomegranates during postharvest storage. The GA-ANN and ANFIS were fed with two inputs of methyl jasmonate and storage time. The results showed that GA-ANN predictions agreed with experimental data and the GA-ANN with 14 neurons in one hidden layer can predict physiological changes and quality parameters of pomegranate (weight loss, pH, chilling injury index, ion leakage, ethylene, respiration, polyphenols, anthocyanins, and total antioxidant activity) with correlation coefficients equal to 0.87. The ANFIS model was trained by a hybrid method and agreement between experimental data and ANFIS predictions was significant (r = 0.90).