摘要
传统除草机器人作业系统对于杂草的辨识不够精准,导致杂草去除召回率、误检抑制率较低。为此设计基于机器视觉的寒地水稻田间除草机器人精准作业系统。设计由图像处理器、摄像头等构成的机器视觉模块,采集寒地水稻田间图像。在寒地水稻田间杂草识别模块中,通过YOLOv3卷积神经网络模型实现采集的寒地水稻田间图像中的杂草识别。采用全电动四轮驱动底盘作为除草机器人的移动装置,基于杂草识别预测框实现除草路径导航。为除草机器人配备灵活的机械臂,搭载激光发射器,基于杂草识别预测框实现除草作业。实例测试结果表明,设计系统能够在寒地水稻田间实现较为精准的除草机器人作业,完成大部分杂草的清除工作,其残留的杂草较少,邻株误伤情况也较少;设计系统的杂草去除召回率整体高于0.9,说明系统对寒地小目标杂草的漏检率低,适应性强;设计系统的误检抑制率整体高于0.85,说明系统对水稻与杂草的形态差异区分能力强,能够减少误除草现象。
Traditional weed control robot operating systems are not precise enough in identifying weeds,resulting in low weed removal recall rates and false detection inhibition rates.Design a precise operation system for weeding robots in cold rice fields based on machine vision.Design a machine vision module consisting of image processors,cameras,etc.to capture images of rice fields in cold regions.In the weed recognition module of cold rice fields,the YOLOv3 convolutional neural network model is used to recognize weeds in the collected cold rice field images.Adopting a fully electric four-wheel drive chassis as the mobile device of the weed control robot,weed path navigation is achieved based on weed recognition prediction boxes.Equip the weed control robot with a flexible robotic arm,equipped with a laser emitter,and implement weed control operations based on weed recognition prediction boxes.The case test results show that the designed system can achieve relatively accurate weed control robot operations in cold rice fields,complete most of the weed removal work,and have fewer residual weeds and fewer incidents of neighboring plant damage;The overall weed removal recall rate of the design system is higher than 0.9,indicating that the system has a low missed detection rate for small target weeds in cold areas and strong adaptability;The overall false detection inhibition rate of the design system is higher than 0.85,indicating that the system has strong ability to distinguish the morphological differences between rice and weeds,and can reduce the phenomenon of false weed control.
作者
吉毅
王娟
JI Yi;WANG Juan(School of Intelligent Manufacturing,Wuchang Institute of Technology,Wuhan430065,China;College of Creative Design,Wuchang University of Technology,Wuhan,Hubei430223,China)
出处
《北方水稻》
2025年第6期183-187,共5页
Northern Rice
基金
湖北省自然科学基金面上项目(2023AFB1107)
湖北省高等教育学会教育科研项目(2023XD109)
中国民办教育协会2025年度规划项目(CANQN250123)
武昌工学院2025年度校级科学研究项目(2025KY02)。
关键词
机器视觉
寒地水稻田间除草机器人
精准作业
YOLOv3卷积神经网络模型
全电动四轮驱动底盘
Machine vision
Weeding robot for rice fields in cold regions
Accurate homework
YOLOv3 convolutional neural network model
Fully electric four-wheel drive chassis
