温室钵苗移栽机的创新设计与钵苗健康信息的视觉检测方法研究

Focuses on essential questions which restricts the development of domestic seedling transplanting robot in greenhouse, the research main in the following two aspects- - innovation design of seedling transplanter and detection methods of healthy information by machine vision in greenhouse.the research is the fundamental technology of seedling automatic transplanting in seedling cultivating for commercial in our country. In this study, we analyze the the development of seedling transplanting equipment in greenhouse in domestic and overseas. Using the main crop as the research object,taking the advantage of multidisciplinary, independent innovation,and international cooperation,we do the research from the system angle. Study on physical and physiological characteristics of seedlings which meet the requirements of mechanical operation, innovation design and optimization of seedling end-effector, detection methods of healthy information of seedlings in plug tray by machine vision and mechanism innovation, planning seedlings transplanting strategy and optimization by algorithm,and so on. Combination the information of seedlings' physical and physiological characteristic information and experimental investigation, do the task of verifing Kinematics and dynamics of end-effector for optimization by the methods of high-speed photography and virtual designing, and do innovation design and optimization of the key mechanism base on experimental. Base on morphology, image segmentation and mechanism innovation designing, we resolve the problem which is difficult to getting the seedling informations of overlapping and cross-border leaves. The project would obtain a breakthrough in theory and key technology for our country in the processing of developing the seedling tranaplanting equipment in greenhouse.

围绕限制国内温室钵苗自动化移栽装备研发中的瓶颈问题，重点开展温室钵苗移栽机关键部件的创新设计与钵苗健康信息的视觉检测方法研究，是开发适合我国温室商品苗实际的钵苗移栽作业装备的基础。本研究在分析国内外温室钵苗机械化移栽现状的基础上，以温室主要栽培作物为研究对象，利用多学科技术的交叉研究、自主创新和国际合作，从系统角度出发，重点开展温室钵苗机械化移栽作业的钵苗可移栽状态、移栽末端执行器的创新设计和优化、适合钵苗健康信息检测的视觉方法和机构创新、移栽作业路径控制策略和优化等研究。结合钵苗物理与生理特性信息和试验研究，利用高速摄影、虚拟仿真现代设计方法等进行末端执行器运动学、动力学分析和仿真，结合试验开展关键机构的创新设计与优化；基于改进的图像处理算法和辅助图像识别的机构创新相结合，解决因叶片重叠、越界等引起的钵苗信息获取难的问题，项目有望获得我国温室钵苗移栽机开发过程中的理论和关键技术突破。

针对项目申请时提出的关键问题，以温室蔬菜为研究对象，陆续开展了各项研究。设计了一种指针可调式夹持压缩苗坨装夹台，并与万能试验机平台组合构成测力系统，研究苗坨在指针夹持下的力学特性，对比试验发现指针夹持角度、夹持指针数、苗坨含水率、根系长度和苗坨基质配比5个单因素对苗坨受指针夹持过程中的抗压力均有影响。为解决钵苗抓取和释放效果不稳定的问题，设计出一种双气缸驱动的钵苗移栽机插入夹取式末端执行器。为提高插入夹取式末端执行器的作业速率，又设计了一种单气缸驱动的插拔夹取式末端执行器，可在气缸的一个推程内连续完成插入和夹取两个动作，且对育苗基质条件、作物品种和穴盘规格都具有较强的适应性，各类测试条件下的移栽成功率均达到100%。基于改进型分水岭算法对跨界重叠不是很严重的穴盘苗叶片进行分割，判别准确率可达到98%以上。若跨界重叠叶片面积过大，算法的判别准确率会大幅下降，为此设计了一种辅助机器视觉识别穴盘苗健康状况的间苗机构，判别准确率达到100%。采用高光谱和荧光成像技术初步实现了番茄黄化曲叶病的早期检测，判别准确率在70.6% 至 96.5%之间。基于智能算法进行了移栽作业的路径规划，大大缩短了移栽作业时末端执行器行走的路径长度，大大提高作业速率。遗传算法和蚁群算法的平均运行时间分别为0.32 s和0.94 s，但是遗传算法不能够处理只有1个空穴的情况，综合比较，蚁群算法更适合用于补苗移栽的路径优化。上述研究不仅提高了移栽机的作业效率，还为相关装置的设计提供了科学依据，对我国温室钵苗移栽机械的发展具有重要意义。