摘要
针对传统的自动导引运输车(automated guided vehicle,AGV)调度方式在半自动生产车间不能满足当前实际生产需求等问题,本文以YOLOv5m为目标检测算法,基于全局视觉构建环境电子地图与路径规划,建立一种基于全局视觉的实时AGV手势调度系统。该系统在全局视觉下通过实时检测工作人员的手势,将手势信息传达给系统信息处理模块,经过解释器处理后,根据不同的手势向AGV发送相应的命令。同时,为验证目标检测的准确率,在全局视觉环境下,基于YOLOv5算法和PyTorch,建立深度学习框架训练目标检测模型,并进行实验验证。实验结果表明,AGV的识别准确率为99.9%,手势的识别准确率为99.7%,且二者检测的置信度均处于0.91~0.99之间,说明模型的检测速度符合实际要求。该系统节省了人力,提高了AGV调度效率,功能和实时性均满足车间实际生产需求,具备良好的稳定性。该研究实现了车间AGV的智能化手势调度,具有一定的实际应用价值。
Aiming at the problems that the traditional automatic guided vehicle(AGV)scheduling method cannot meet the current actual production demand in the semi-automatic production workshop,this paper focuses on the design of the workshop AGV gesture scheduling system based on global vision.With YOLOv5m as the target detection algorithm,we construct an electronic map of the environment and path planning based on global vision,and at the same time establish a real-time AGV gesture dispatching system based on global vision,which detects the staff′s gestures in real time under global vision,communicates the gesture information to the system information processing module,and sends corresponding commands to the AGV according to different gestures after the interpreter processing.Meanwhile,to verify the accuracy of target detection,a deep learning framework is established to train the target detection model based on YOLOv5 algorithm andPyTorch under global vision environment,and experimental verification is conducted.The experimentalresults show that the recognition accuracy of AGV is 99.9%and the recognition accuracy of gesture is 99.7%,and the confidence of both detection is between 0.91 and 0.99,indicating that the detectionspeed of the model meets the practical requirements.The system saves manpower,improves the efficiency of AGV dispatching,its function and real-time meet the actual production requirements of the workshop,and it has good stability.The study realizes intelligent gesture dispatching of workshop AGVsand has practical application value.
作者
李晓帆
刘泽平
陈世海
麻方达
姚明杰
符朝兴
LI Xiaofan;LIU Zeping;CHEN Shihai;MA Fangda;YAO Mingjie;FU Chaoxing(College of Mechanical and Electrical Engineering,Qingdao University,Qingdao 266071,China;Qindao Baojia Intelligent Equipment Co.,Ltd.,Qingdao 266113,China)
出处
《青岛大学学报（工程技术版）》
CAS
2023年第1期34-42,共9页
Journal of Qingdao University(Engineering & Technology Edition)
关键词
AGV调度
人体手势
目标检测
全局视觉
实时
路径规划
AGV scheduling
human gesture
object detection
global vision
real time
path planning
