Research And Design Of Automatic Docking System Based On Machine Vision

With the upgrading of China’s technological transformation of traditional enterprises,machine vision used to help people find out what’s going on.Device vision has the advantage of a touchless,flexible and efficient operation,which can meet the needs of various industries Based on the work done by a team of magnesium industry as an application background,this paper studies and manufactures an auto-vision-based system with the aim of auto-positioning between the guide cylinder and the reduction tank mouth in the automatic feeding process.In view of the characteristics of this application scenario,this paper combines object detection with machine vision guidance system,comprehensively considers the alignment accuracy,time constraint,docking success rate and other factors,and analyze the target detection method and the workspace based on the vision of the machine,and finally realize the accurate automatic docking between the metal magnesium feeding cart and the reduction tank.The main work is as follows:First,the construction process is carried out According to the real needs the target identification and positioning part of the automatic docking reduction tank in the project,the design of the building has been with a software program such as the core,the scheme designs the functional modules of each subsystem and the interaction mode and communication interface between each subsystem.Next,the selection and positioning method of the target detection method were studied.Based on the requirements of the project,the docking process is divided into two parts of coarse positioning and precise positioning,to study the RFID technology based on the Internet for coarse positioning,to guide the loading cart to a specified reduction tank position,and to realize coarse positioning.In order to realize accurate positioning,the target detection and positioning technique based on the depth learning is studied,and the target detection and positioning of the reduction tank mouth are realized using yolov3 Ti Ni.Secondly,the methods of object detection and localization were studied.According to the application scenario,the automatic docking process is divided into two parts: coarse positioning and fine positioning.For coarse positioning,a coarse positioning scheme based on RFID technology is given,and the feeder is guided to reach the designated target position by continuously reading the electronic tag information through the card reader,so as to achieve coarse positioning.For precise positioning,the target detection and localization technology based on deep learning is studied,and YOLOv3-Tiny is used to realize the target detection and positioning of the reduced tank mouth.The actual test results show that YOLOv3-Tiny has the characteristics of high precision and good real-time performance,which can meet the detection and positioning needs of the reduced tank mouth.Third,different from the method of calculating the spatial coordinates of the center point through the camera model and target edge extraction,this paper gives the method of calibrating the center point coordinates of the target by pre-training in the offline stage,so as to give the coordinates of the docking center point in the image plane.Aiming at the problem that the reduced tank mouth is not in one plane,a method of real-time dynamic correction of the coordinates of the center point is given,which further improves the success rate of docking.The system has finally been used and tested.The results showed that in the fully automatic operation mode,the docking success rate is more than 90%,which can meet the needs of industrial production.