Research On The Key Issues Of Automatic Resin Eyeglasses Casting And Identification System Based On Machine Vision

With the rise of artificial intelligence,big data,industrial Internet and other technologies,the traditional manufacturing industry can no longer meet the actual production needs,and China’s manufacturing industry is developing in the direction of automation and intelligence.Based on the resin eyeglass lens manufacturing industry,this study focuses on solving the fully automated production of resin eyeglass lenses,completely liberating the manual labor and improving the production efficiency and production quality of enterprises.This paper proposes a machine vision-based automatic resin lens casting method,which innovatively integrates the lens sorting process with the automatic casting system,and achieves high precision automatic resin lens casting while completing the recognition of the cast lens degree characters,facilitating the subsequent sorting of the lenses and counting the production output.This paper focuses on the detailed study of the vision algorithm involved,and the specific work is as follows:(1)For the problem of detection of degree characters on the mold surface,this study proposes a character detection algorithm based on deep learning,and after detailed experimental comparison,the DBNet model with ResNet18 as the feature extraction network is selected,and further improvements are made on the basis of the basic DBNet model,and finally the accuracy of the improved DBNet model reaches 93.1% and the recall rate reaches82.3%,-(80)index reached 87.4%.(2)For the problem of mold surface degree character recognition,this study extracts CRNN+CTC as the base model for character recognition.In order to improve the recognition accuracy of the algorithm on the data set of this study,this study improves the base CRNN structure accordingly,and the recognition accuracy of the improved model reaches 93.6%after experimental comparison.Finally,in the task of performing end-to-end lens degree recognition in industrial scenarios,the accuracy of the improved DBNet as the character detection model and the improved CRNN as the character recognition model proposed in this study reached 84.2% and the frame rate reached 10 frames per second.(3)For the study of automatic casting algorithm,this paper innovatively adopts a dualcamera combined visual inspection scheme,which measures the thickness of the mold gap and the height of the liquid level line inside the mold by the dual cameras installed in the radial and axial directions of the mold respectively,so as to calculate the real-time cavity volume and realize the high-precision quantitative resin casting with the flow pump.The algorithms of mold gap dimension measurement,mold area positioning and liquid level detection were further studied in detail,and the designed vision algorithms were integrated into the vision software for experimental testing.The experimental results showed that the automated casting system designed in this research can meet the actual production requirements in terms of casting speed and casting quality.