Research On Products Recognition And Location Based On Machine Vision

In order to develop the domestic manufacturing industry,the world's manufacturing powers have put forward the strategy of revitalizing the domestic manufacturing industry.From "industry 4.0","industrial Internet" to "made in China 2025",its essence is to promote the technology research,practical application and industrial upgrading of domestic intelligent manufacturing.In the background of intelligent manufacturing,manufacturing enterprises are required to carry out digital and automatic transformation of factories.As a new technology,machine vision has made great contribution to the unmanned,automatic and intelligent manufacturing enterprises.The wide application of machine vision technology in manufacturing industry is the inevitable result of the reform of manufacturing industry.In view of this,this paper makes an in-depth study on the recognition and location methods of the products in the production line by machine vision in the intelligent manufacturing environment,mainly including: Research on image acquisition and data expansion methods,Research on product recognition methods based on deep learning,Research on product location methods based on the Yolo detection framework(also known as product detection methods),and application of products recognition and location methods verification.Firstly,this paper studies the image acquisition and data expansion methods of the product line,and improves the generalization ability of the image recognition model by enriching the diversity of training samples.Then,through the study and research of Convolutional Neural Network,the products recognition model is built,and the model parameters are optimized,the recognition accuracy is 97.5%.In order to meet the real-time requirement of the production line for product detection,the Yolo detection framework is selected comprehensively.On the basis of it,the convolution structure of the model and the anchor point are added to make it meet the requirements of product identification and positioning of the production line.Finally,the application verification of the products recognition and location method of the production line is carried out.Through the complete method application process,including the selection of camera,lens and light source,the construction of software environment and model construction,the application verification of the identification and positioning method is completed.The detection speed of the model is kept at 13 fps on the premise of ensuring the accuracy,which can meet the demand of real-time.