摘要
生产线生产过程中需要回收产品标签上的信息,确认产品去向并核对。针对人工回收标签信息不仅浪费人力,并且失误率高的问题,文中提出了一种基于轻量化Ghost模块改进的YOLOv4目标检测算法,用于检测生产线上产品的标签。通过连通域分割方法将标签上的点阵字符串分割成单个字符,通过卷积神经网络进行点阵字符的识别,回传到系统中进行信息的回收和核对。文中通过工业相机以及传送带为实验平台模拟药瓶出厂试验,得出检测速度相较于传统YOLOv4目标检测算法提升了133%,信息录入准确率为99.87%的结果,具有很好实际应用价值。
In the production process,the production line needs to recycle the information on the product label,confirm the whereabouts of the product and check.In view of the problem that manual recycling label information is not only a waste of manpower,but also a high error rate,this paper proposes a modified YOLOv4 target detection algorithm based on lightweight Ghost module to detect the label of products on the production line.The lattice string on the label is divided into a single character through the connected domain segmentation method,and the lattice characters are identified through the convolutional neural network,and sent back to the system for information recovery and verification.This paper simulates the industrial camera and conveyor belt as the experimental platform,and shows that the detection speed is 133%higher with the traditional YOLOv4 target detection algorithm,and the information input accuracy is 99.87%,which has good practical application value.
作者
王晨光
WANG Chenguang(School of Information Science and Technology,Qingdao University of Science and Technology,Qingdao 266061,China)
出处
《电子设计工程》
2023年第12期1-5,共5页
Electronic Design Engineering
基金
国家重点研发计划(2018YFB1702902)。
关键词
生产线
工业视觉
目标检测
字符识别
product line
industrial visual
object detection
character recognition
