LED TV Screen Inspection using Deep Learning toward Machine Vision

Abstract

For the purpose of implementation on the whole process of the industrial 4.0 era with respect to intelligent manufacturing of electronic enterprises, machine vision is introduced to complete the product defect detection, which is used to solve the diversity problem of deficiency on defects in the algorithm abilities, and further to improve the product quality and efficiency by the industrial robot in the automatic production process. First of all, the feature model is obtained through deep learning for the small sample defect dataset. Then the transfer learning method is applied to implement the detection of LED TV through the trained feature model, and the detected production is stored as a sample data to further adjust the model parameters by incremental learning method, and finally the FCNet (Fully Connected Neural Network) is used to complete the classification. Some optical screen detection technology for LED TV oriented machine vision is discussed, and the incremental learning model of the sample dataset is given as a continuous supplement. The full measurement and analysis of screen using depth learning algorithms for different detection targets and sizes are carried out. Experiments show that using the deep learning technology will further improve the accuracy of automatic detection, especially in the field of intelligent manufacturing using industrial robots, will enhance the flexibility and intelligence level of industrial production, and to expand the application of machine vision to provide demonstration.