Research And Implementation Of Face Ethnicity Features Extraction And Recognition Technology

The automatic extraction of facial features to verify the identity of a person has been widely used in various fields,but the effect of face recognition depends on the diversity of people and has great volatility.Since China is a multi-ethnic country,studying the facial features of different ethnic groups can not only promote the current face recognition and detection,but also preserve this feature through modern information technology means,which is conducive to the study of ethnic integration and reproduction.The new features provide valuable research materials for ethnology and anthropdogy.With the paper published by Hinton in 2006,the deep neural network has achieved great results both in theory and in practical applications,and is applied in all aspects of our lives.However,retraining a deep neural network from the beginning requires a lot of computational resources.The emergence of transfer learning has greatly reduced the application threshold of deep neural networks,and many fields can enjoy the dividend brought by AI development.This paper first builds a face database dedicated to ethnic recognition.Because the difficulty of collecting a large number of ethnic faces in real life is too great,the data collection is mainly done by web crawlers.Then based on the processed data,the convolutional neural network dedicated to face recognition is realized firstly,and then the performance of ethnic recognition of different pre-trained networks for face is compared by migration learning.The experiment is verified that transfer learning is more efficient.On this basis,the object detection network dedicated to the nation is constructed and analyzed.Finally,the face nationality video recognition system based on the above model is designed and implemented,and achieves good results.