Research And Implementation Of Intelligent Monitoring System For Product Assembly

Product assembly is an important link in the delivery of industrial products.Products with complex assembly processes can only be assembled by workers.The problem of missing parts often occurs during the assembly of complex products.The traditional reliance on human supervision of the assembly process has limited effect.The intelligent monitoring system for product assembly can replace manpower for monitoring,to save human resources and improve product yield.At present,the intelligent monitoring system of product assembly in the industry usually needs to collect a large amount of data in the actual assembly scene in advance.The system uses professionals to label the data and trains the model on the obtained data set.In the face of product style update iterations,the previous model has limited monitoring effect in new products.People usually need to re-collect the data for data labeling.The slow speed of human labeling and the lack of professional labelers in the factory lead to a slow iteration of the existing system model,which cannot meet the industrial scenario of rapid product iteration.This thesis aims to study an intelligent monitoring system for product assembly,which can intelligently monitor the product assembly process and enable the model to iterate rapidly with the product.In this thesis,we have systematically implemented the intelligent monitoring of the assembly of tent products.The main work of this thesis is as follows:1.We propose and implement data annotation and data augmentation methods.First,according to the image data of the product,we innovatively uses the data annotation method based on similarity comparison to get the annotation.Then we use image synthesis,horizontal flip,brightness enhancement and other operations to perform data enhancement on the labeled data to obtain a dataset.Compared with the manual labeling method,this method improves the labeling efficiency and solves the problem that the model update speed of the intelligent monitoring system is slow in the face of product iteration.2.We propose a complete set of intelligent monitoring solutions.First,we perform data annotation and data enhancement based on product basic data to obtain a data set.Second,we train the object detection model YOLOv4-tiny on the dataset and deploy the model.Finally,the camera equipment is used to collect the data in the assembly scene in real time.The data is produced to the Kaflca.We use the YOLOv4-tiny model to predict the assembly data.When encountering an error in the assembly process,the user is notified in real time to realize intelligent monitoring.3.We have researched and implemented an intelligent monitoring system for product assembly.This thesis combines data annotation methods,data enhancement methods and intelligent monitoring schemes.For the specific industrial scenario of tent product assembly,we have implemented the system.First of all,we have carried out requirement analysis,function analysis and outline design of the system.Then we divide the system into data acquisition module,camera service module,tent service module,model service module and intelligent monitoring module.Finally,we design and implement each module in detail.