结合YOLOv5与双目视觉定位的摘果机械臂控制研究  

Research on the Control of Fruit Picking Manipulator Based on the Fusion of YOLOv5s and Binocular Vision

作　　者：李俊萩 王甲一 孔德肖 张晴晖 Li Junqiu;Wang Jiayi;Kong Dexiao;Zhang Qinghui(College of Big Data and Intelligent Engineering,Southwest Forestry University,Kunming 650224,China)

机构地区：[1]西南林业大学大数据与智能工程学院,昆明650224

出　　处：《农机化研究》2024年第7期38-43,共6页Journal of Agricultural Mechanization Research

基　　金：云南省科技厅重大专项生物资源数字化开发应用项目(202002AA10007);云南省教育厅科学研究基金项目(2021J0153)。

摘　　要：为满足果实采摘需求,设计了一种基于YOLOv5与双目视觉定位的采摘机器人。通过一个全向轮底盘搭载六自由度机械臂,利用YOLOv5目标检测算法实现果实种类识别及平面坐标测量;利用双目视觉实现深度坐标测量,将两者相融合并转换得到机械臂坐标系下的果实三维坐标x、y、z,由三维坐标进行逆运动学解算得到机械臂各个关节角度,最终通过控制关节角度的改变实现果实抓取,并结合Android、物联网技术实现实时视频远程查看与机器人远程控制。试验结果表明:基于YOLOv5实现3种果实识别模型的训练,模型精确率99.7%,召回率99.8%,mAP(0.95)89.03%,每种水果的识别准确率平均为95%;机械臂投射到二维平面进行逆运动学解算,降低了运算量与运算时间,平均解算时间为0.252s。分别对仿真桃子、桔子、石榴进行采摘50次,平均单次采摘时间为13.48、13.91、13.76s。系统具有良好的实时性与准确性,适宜布署于嵌入式系统,满足果实采摘需要。In order to meet the demand of fruit picking,this paper designs a picking robot based on YOLOv5 and binocu-lar vision.A six-degree-of-freedom robot arm is mounted on an omnidirectional wheel chassis,and the YOLOv5 target detection algorithm is used to identify fruit species and measure plane coordinates,while binocular vision is used to meas-ure depth coordinates.Finally,the fruit is grasped by controlling the change of joint angle.Combined with Android and IoT technology,we realize real-time video remote viewing and robot remote control.Based on YOLOv5,the training of three fruit recognition models was realized,and the model accuracy was 99.7%,recall rate was 99.8%,mAP(0.95)was 89.03%,and the recognition accuracy of each fruit was up to 95%on average;the robot arm projected to the two-dimensional plane for inverse kinematic solving,which reduced the operation volume and operation time,and the average solving time was 0.252s;the peach,orange,and pomegranate models were picked separately.The average single picking time is 13.48s,13.91s and 13.76s.The system has good real-time and accuracy,and can meet the fruit picking re-quirements.

关 键 词：YOLOv5 双目视觉 机械臂 逆运动学 果实采摘 

分 类 号：S225.93[农业科学—农业机械化工程] TP391.4[农业科学—农业工程]