奇异果采摘机器人机械臂控制系统研究——基于计算机视觉和卷积神经网络  

Research on the Control System of the Manipulator of Kiwi Fruit Picking Robot——Based on Computer Vision and Convolutional Neural Network

作　　者：冯硕[1] 李旗 Feng Shuo;Li Qi(Henan Polytechnic Institute,Nanyang 473000,China;Henan Material Forming Equipment Intelligent Technology Engineering Research Center,Nanyang 473000,China)

机构地区：[1]河南工业职业技术学院,河南南阳473000 [2]河南省材料成形装备智能技术工程研究中心,河南南阳473000

出　　处：《农机化研究》2022年第10期25-29,共5页Journal of Agricultural Mechanization Research

基　　金：河南省科技发展计划项目(182102210266);河南省科技攻关项目(192102210165)。

摘　　要：介绍了计算机视觉的架构和原理,基于卷积神经网络设计了奇异果的目标检测和识别方法,并利用双目视觉原理实现了奇异果立体定位,使得系统能够实现目标果实的立体定位和精准采摘。实验结果表明:采摘机器人机械臂可以根据控制指令准确达到目标位置进行采摘,实际到达的位置与视觉定位坐标误差在4.5%以内,能够满足采摘机器人采摘要求。It first introduces the framework and principle of computer vision,then designs and studies the method of target detection and recognition of kiwi fruit based on convolution neural network,and realizes the stereo positioning of kiwi fruit by using the principle of binocular vision,so that the system can achieve the target fruit positioning and precise picking.The experimental results show that the mechanical arm of the picking robot can accurately reach the target position according to the control command,and the error between the actual position and the visual positioning coordinate is within 4.5%,which can meet the picking requirements of the picking robot,which proves that the picking manipulator has high accuracy.

关 键 词：奇异果 采摘机器人 计算机视觉 卷积神经网络 

分 类 号：S225[农业科学—农业机械化工程]