Button Battery Sorting System Based On Machine Vision And Parallel Robot

As our daily necessities,button batteries have massive consumption and production volume every day.In the production process of button batteries,the inspection of appearance defects is a strict and necessary process.However,this process is still manually picked by the naked eye.Due to the large amount of repetitive work,the visual fatigue of the staff will be caused.Missing and wrong inspections of defective products often occur,which will inevitably cause product quality problems for enterprises.This paper designs a button battery sorting system based on machine vision and parallel robots to realize this process’ s automatic sorting.The following is the content involved in this paper:(1)The general introduction of the system,according to the enterprise’s problems,designed the entire system’s framework.The system mainly consisted of three subsystems:machine vision subsystem,mechanical and electronic control subsystem,and parallel robot subsystem.Each subsystem is independent and needs to be equipped with the required hardware and software,and at the same time,cooperate with each other.During the actual operation of the system,only by completing accurate communication between the three subsystems,can the correct sorting of button batteries be realized.(2)Image algorithm research,this part is the main content of this article.According to the two types of images collected by the machine vision subsystem,two image algorithms are used.One is the image collected for defects such as depressions.This article uses the YOLO target detection algorithm based on convolutional neural network to locate the depression and improve the YOLO algorithm’s loss function.Using CIOU as the target frame positioning loss and replace the Focal Loss with the cross-entropy loss to calculate the confidence value.In the experiment,it can be seen that the m AP of improved model is74.76%,an increase of 0.84%.Two is for images collected with other defects.This article uses an improved lightweight network,Mobile Net V2,to classify a single button battery.To further reduce the calculation of Mobile Net V2 while maintaining recognition accuracy,adding channel shuffle to the basic module of Mobile Net V2 is proposed,the final test set accuracy rate is 98.31%,which is an increase of 3%,and the time complexity is reduced by33%.(3)The design of machine vision software,the machine vision software mainly receives the image data sent by the industrial smart camera,integrates the image algorithm,processes the collected images,and then sends the detection results to the parallel robot to sort the button batteries.Therefore,the design of machine vision software mainly includes four modules: a software interface module,an image acquisition module,an image processing module,and a communication protocol module.(4)System testing,system testing mainly includes machine vision software testing,parallel robot linkage testing,and system performance testing.After testing,the machine vision software can run continuously for a long time,the parallel robot can also accurately and stably realize the sorting function,and the system performance also meets the actual needs.The removal and completion time of a single button battery is within 2.5s,and the defect identification of the button battery The rate can reach 98.15%.Aiming at the shortcomings in button batteries’ actual production process,this paper proposes a button battery sorting system based on machine vision and parallel robots.Image algorithms and machine vision software are the main research content in this system,and did some system test.According to the test results,it is shown that the system can meet actual application requirements,has specific commercial value,and can be promoted and used.