Research On Face Modeling Algorithms Based On Expression Classification And Feature Extraction

Three-dimensional face modeling technology can be widely used in film,animation,games,interior design and medical cosmetology and other industries.In the past research,there are some problems in the three-dimensional face model obtained by traditional modeling technology,such as the modeling effect is not realistic enough and can not meet the real-time requirements.Aiming at this problem,this paper proposes a method of 3D face modeling based on the combination of expression classification and feature extraction.Firstly,the face in the image is located and the feature point information is extracted.These feature points mark the facial contour,eyebrow,eye contour,bridge of nose,next circumference of nostril and mouth contour respectively.Then CNN algorithm is used to classify the displacement degree of feature points,so as to extract feature points and classify facial expressions.A subdivision algorithm of Candide-3 facial model based on different region granularity is proposed.The facial expression model library is established for the position of the feature points of Candide-3 model after subdivision,and the corresponding Candide-3 facial model is invoked according to the result of facial expression classification.Finally,the expression model is further refined by RBF interpolation algorithm.3D facial expression animation system.Through the design and implementation of the above algorithm ideas,the development and detection of face modeling animation system is completed.In the detection experiment,for the acquired video image,the convolutional neural network is used to classify the facial expressions after extracting the feature points,and then the personalized adjustment of the three-dimensional facial model is made based on the three-dimensional facial model.Finally,the corresponding three-dimensional facial model is obtained.The experimental results show that the design of the proposed method makes the head of the face model more complete,the information of face and contour more natural,and meets the requirements of real-time modeling.