【作者】 李玉华；

【导师】 侯加林；

【作者基本信息】 山东农业大学 ， 农业电气化与自动化， 2021， 博士

【摘要】 大蒜作为重要的调味品,具有丰富的营养价值和保健及药用价值,深受大众喜爱,种植面积逐年增加。中国作为最大的生产和消费国,据统计种植面积已超过90万hm~2。机械化种植已成为必然趋势,但由于大蒜播种要求“单粒播种、根下芽上、直立定深栽种”的严格种植农艺,国内大蒜机械化播种水平还处于较低水平。目前,阻碍国内大蒜机械化播种推广的主要原因有:现有大蒜播种机存在单粒取种合格率低、种芽朝向调整合格率低且适应性差和定向插播直立度偏低的问题。本研究针对大蒜机械化播种存在的问题,以大蒜为研究对象,从大蒜三维尺寸的数理统计入手,借助离散元仿真技术和数学规划模型对轮勺式单粒取种装置的关键结构及参数进行优化,对提高轮勺式单粒取种装置单粒取种合格率的方法进行了研究;通过对大蒜图像分析提取了3个大蒜种芽朝向特征参数,借助Lab VIEW平台搭建了种芽朝向调整试验台对推动式种芽朝向调整方法进行了验证;对利用边缘计算处理器Jetson Nano和卷积神经网络实现种芽朝向调整进行了研究,解决了嵌入处理器计算能力不足而工控机体积偏大的问题,为实现智能化种芽朝向调整装置在产品端的部署提供了参考;通过离散元仿真分析结合台架试验从微观角度对大蒜—插播器—土壤三者之间交互作用的机理进行了分析,为提高大蒜种植的直立度提供了理论依据。主要研究内容归纳如下:(1)轮勺式大蒜单粒取种装置研究。对取种区、输种区和排种区的大蒜进行了受力分析,阐述了轮勺式大蒜单粒取种的机理。通过离散元仿真分析对不同取种勺及取种轮的结构形状进行了对比优化,确定了取种勺及取种轮最优结构。以大蒜三维尺寸的正态分布数据为基础,建立了取种勺关键尺寸的数学规划模型,通过图解法确定了取种勺实现单粒取种的尺寸区间。在理论分析的基础上搭建了轮勺式单粒取种试验台,以取种勺的半径、长度和取种轮转速为试验因素,以漏充率和合格率为响应指标进行了正交回归试验,建立了漏充率和合格率的回归模型,并对回归模型进行了参数优化。(2)基于机器视觉的大蒜种芽朝向调整装置设计与试验。在对大蒜图像分析的基础上,提出了三个与种芽朝向极其相关的三个特征参数,并利用多特征参数融合的方法获得识别大蒜种芽朝向的综合特征参数,利用MATLAB对该种芽朝向识别算法的正确率和适应性进行了试验验证。种芽朝向识别算法研究的基础上,设计了基于机器视觉的横向推动式大蒜种芽朝向调整装置,对该装置的种芽朝向调整机理进行了分析,借助Lab VIEW平台对该装置的硬件系统和软件系统进行了设计。搭建了试验台对该装置的种芽调整效果进行了验证。(3)基于Jetson Nano的大蒜种芽朝向调整装置设计与试验。创新设计了料斗翻转式大蒜种芽朝向自动调整装置,对该装置的种芽朝向调整机理进行了分析。建立了种芽朝向调整离散元模型,通过仿真分析结合正交试验对该装置的关键结构参数进行了优化。控制算法方面,为降低模型复杂度提高种芽朝向判断实时性,采用双卷积神经网络模型结构,分别用于大蒜喂入检测和种芽朝向判断。控制系统以NVIDIA的边缘计算处理器Jetson Nano为核心,借助NVIDIA Tensor RT加速库对双卷积神经网络模型进行推理,解决了控制器计算能力与体积不能兼顾的问题。以杂交大蒜为试验对象对该装置进行了台架试验验证。(4)大蒜-插播器-土壤交互的定向插播作用机理分析与试验。对双鸭嘴型定向插播器的工作机理进行了分析,建立了大蒜-插播器-土壤相互作用的离散元模型。分别对间歇插播和连续插播两种作业形式的工作过程进行了仿真分析,揭示了插播过程大蒜、插播器与土壤三者之间的相互作用机理。分析了插播深度、提升高度、插播速比和土壤颗粒表面能对插播直立度及种植深度一致性的影响。搭建了基于Labview的试验台架对分析结果进行了试验验证。仿真及试验结果表明:(1)插播深度、提升高度及插播速比均对种植直立度有直接影响,这些参数只有相互协调且取值合理才能有效提高种植直立度;(2)土壤颗粒表面能对种植直立度有较大的影响,表面能越小直立度越高。更多还原

【Abstract】 As an important condiment with rich nutritional value and health and medicinal value,garlic is popular among the public,and the planting area increases year by year.China is the largest producer and consumer of garlic,with a garlic planting area of more than 900,000hm~2 according to statistics.Mechanized sowing has become an inevitable trend,but the level of mechanized garlic Mechanized sowing in China is still at a low level,due to the strict agronomic requirements of garlic planting,i.e.,"single grain sowing,root down and bud up,upright and planting depth consistent".The existing garlic sowing machines have a low qualification rate of single grain seeding,low qualification rate,poor adaptability of garlic clove orientation adjustment,and low uprightness in directional sowing.These are the main factors that hinder the promotion of mechanized garlic sowing in China.This study took garlic as the research object and aimed to solve the problems in garlic seeder.The key structure and parameters of the wheel and spoon single-seed extraction device were optimized with the help of discrete element simulation technology and a mathematical model.Also,the method of improving the qualification rate of single-seed extraction by wheel and spoon single-seed extraction device was studied.Besides,three characteristic parameters of garlic clove orientation were extracted by analyzing garlic images,and a garlic clove orientation adjustment test bench was built with the Lab VIEW platform to verify the effectiveness of the pushing garlic clove orientation adjustment method.The edge computing processor Jetson Nano and convolutional neural network were used to adjust the garlic clove orientation.This solves the problem of the insufficient computing power of the embedded processor and the large size of the industrial control machine and provides a reference for deploying the intelligent garlic clove orientation adjustment device at the product end.Additionally,the mechanism of the interaction between garlic-inserter-soil was analyzed from the microscopic perspective through discrete element simulation analysis and bench test.This provides a theoretical basis for improving the uprightness of garlic planting.The main contributions of this study are summarized as follows.(1)The wheel and spoon single-seed extraction device was studied.Through the stress analysis of garlic in seed taking area,seed conveying area,and seed metering area,the mechanism of single garlic seed taking with wheel spoon was investigated.The optimal structure of the seed scoop and seed-taking wheel was determined by comparison and optimization through discrete element simulation analysis.Based on the normal distribution data of the three-dimensional size of the garlic clove,the mathematical model of the key size of the seed picking scoop was established,and the size range of a single seed picking scoop was determined by the graphic method.Through theoretical analysis,a wheel spoon type garlic single seed collection test-bed was built.The orthogonal regression test was conducted,with the radius and length of the seed picking scoop and the speed of the seed picking wheel as the test factors,and the leakage rate and qualified rate as the response indexes.Besides,the regression model of leakage rate and the qualified rate was established,and the parameters of the model were optimized.(2)The adjustment device based on machine vision for garlic clove direction was designed and tested.By analyzing garlic clove images,three feature parameters that are closely related to garlic clove orientation were proposed,and a multi-feature parameter fusion method was used to obtain the comprehensive feature parameters for recognizing garlic clove orientation.The correctness and adaptability of the garlic clove orientation recognition method were experimentally verified by using MATLAB.Based on the research on the garlic clove orientation recognition method,a machine vision-based lateral push garlic scale bud orientation adjustment device was designed.Meanwhile,the mechanism of garlic clove orientation adjustment was analyzed,and the hardware and software systems of the device were designed on the Lab VIEW platform.Besides,a test bench was built to verify the effectiveness of the device for garlic clove adjustment.(3)The garlic clove orientation adjustment device was designed and tested by using the Jetson Nano processor.The automatic garlic clove orientation adjustment device was designed,and the mechanism of the garlic clove orientation adjustment device was analyzed.Meanwhile,the discrete element model of the garlic clove orientation adjustment device was established,and the key structural parameters of the device were optimized through simulation analysis in combination with orthogonal tests.In terms of the control algorithm,to reduce the model complexity and improve the real-time judgment of garlic orientation,a dual convolutional neural network model was used for garlic clove feed detection and garlic clove orientation judgment.Jetson Nano,an edge computing processor from NVIDIA,was used as the core of the control system,and the NVIDIA Tensor RT acceleration library was used to reason about the convolutional neural network model,reconciling the computing power and volume of the controller.The device was validated in a bench test using a hybrid garlic clove as the test object.(4)The interaction mechanism between garlic clove,inserter,and soil during directional planting was investigated.The working mechanism of double duckbill type directional inserter was analyzed,and the discrete element model of garlic-inserter-soil interaction was established.Based on this,the working process of intermittent insertion and continuous insertion were simulated and analyzed to reveal the interaction mechanism between garlic clove,inserter,and soil during directional planting.The effects of insertion depth,lifting height,insertion speed ratio,and soil particle surface energy on insertion uprightness and planting depth consistency were analyzed.Besides,a testbench was built on Labview to experimentally validate the analysis results.The simulation and test results showed that:(1)insertion depth,lifting height and insertion speed ratio have direct influence on planting uprightness,and these parameters can effectively improve planting uprightness only if they are coordinated with each other and have reasonable values;(2)the surface energy of soil particles has a greater influence on planting uprightness,and the smaller the surface energy the higher the uprightness.更多还原