美国白蛾幼虫网幕对靶喷药机器人弱标定视觉伺服控制研究

Hyphantria cunea is a kind of worldwide quarantine pest, which spread rapidly in china. The existing application device for spraying is low efficiency, large amount of labor and seriously pollute the environment. It is necessary to have research on target spraying robot with high accuary. This project intends to explore the typical broadleaf tree modeling and the formation Law of Hyphantria cunea Larva nets, then establish the training sample library. On this basis，high real-time and robust recognition method of Hyphantria cunea Larva nets will be studied. The method of weak calibration for the information of the Hyphantria cunea Larva nets will be researched to realize the conversion from image space to robot space. A method for determining the optimal spraying position will be researched by the planning of the system’s motion path ,in order to achieve the movement of the shortest distance, the minimum amount of spraying. This project can realize the target spraying of Hyphantria cunea Larva Nets, improve the spraying efficiency, reduce pesticide pollution, so it is of great practical significance and application value .

美国白蛾是世界性检疫害虫，在我国蔓延迅速严重危害林业资源。现有施药方法与装置针对性差、施药效率低，施药劳动量大、污染环境严重，急需研究对靶施药机器人，实现自动化精准喷药。为此，本项目拟对典型阔叶树树形建模，探索美国白蛾幼虫网幕结网规律，建立训练样本库，研究实时性高、鲁棒性强的网幕轮廓快速识别方法；研究施药环境下美国白蛾幼虫网幕轮廓信息的施药机器人弱标定方法，实现图像空间到机器人空间的精确转换；规划手眼机构运动路径，研究一种最优喷药位置判断方法进行实时判断，进而修正手眼机构运动参数，以期达到运动距离最短、喷药量最省的目的，为研制智能对靶施药机器人奠定基础。本项目的研究对于实现美国白蛾幼虫网幕的对靶施药防治，提高施药效率，减轻施药污染，具有重大的现实意义与应用价值。

美国白蛾是一种世界性防疫害虫，其幼虫吐丝结网危害树木，目前常采用人工喷洒农药的防治方式，效率低，污染大。研究智能对靶喷药技术，实现美国白蛾幼虫网幕的精准识别及自动喷药，对于降低劳动强度，改善生态环境，促进可持续发展，具有重大的现实意义。本研究在对美国白蛾幼虫网幕分布特征充分分析的基础之上，开发了一种双闭环弱标定视觉伺服控制系统并研制了空间四自由度喷药机器人机械本体。系统使用基于卷积神经网络与不重合滑动窗口法相结合的识别方法对美国白蛾幼虫网幕进行识别，规划行走路径，实现最佳喷药位置自主判定。经自然光状态下的喷药实验验证，系统的处理速度、处理精度均满足工作要求，无漏喷现象。本研究填补了国内美国白蛾幼虫网幕自动喷药机器人领域的空白，具有良好的应用前景。