Research On Personalized Modeling Technology Of Peking Opera Face Based On Depth Camera

3D personalized modeling technology is a hot research topic in computer vision,virtual reality and other fields.As the quintessence of our country,the Beijing Opera has also attracted the attention of domestic and foreign scholars in recent years,and achieved some research results.At present,there are two main ways to realize the personalized Beijing Opera Face,one is the research of two-dimensional projection of face makeup,the other is the layer model by layer modeling of vectorized face.However,both of the above methods are based on the face pattern vector library,and there are some problems,such as the lack of realism,or the need for manual deformation so that can't do real-time personalization.Therefore,this paper studies the personalization of Peking Opera Face Based on depth camera,in order to make the personalization effect more realistic and real-time driving.The research content mainly includes the registration and splicing of multi frame face point cloud Detection and tracking of facial data and personalized face deformation technology,the specific research contents are summarized as follows:Firstly,the region division of AU(Action Unit)unit of point cloud based on RGB image 3D mapping is designed,and the method of coarse registration of point cloud based on fast point feature histograms(FPFH)of AU unit is proposed.For the low efficiency of the point cloud registration method of the iterative closest point(ICP),the AU local feature is used to replace the global feature of the point cloud to improve the efficiency of feature extraction,and the AU local feature is used for coarse registration to solve the high deviation problem of the ICP point cloud registration method.Because this method only extracts the point cloud features in the representative AU unit,it has higher efficiency and accuracy than the conventional FPFH coarse registration,and can provide a perfect initial transformation matrix,which is conducive to solving the high deviation problem of ICP registration,andprovides a good foundation for the key point feature extraction of personalized face.Secondly,a face detection and tracking method based on LBP(local binary pattern)feature and skin color information is designed.In face detection,the side face is easy to lose and the tracking delay is real-time.We use the skin color information as the additional condition of face tracking.At the same time,we use the LBP vector feature which can express the face texture feature as the training input feature of the Adaboost classifier.The experimental results show that LBP feature is more efficient than Haar feature in face detection,and the method of adding skin color information to alternate tracking can effectively reduce the error of side face loss,which is more conducive to real-time tracking and facial feature point extraction,and provide stable data for the parameter mapping of personalized Peking Opera Face spectrum.At last,the AU element of 3D standard face is divided and the shape base is constructed by local deformation.Based on the 3D standard Peking opera face,this paper studies the 3D deformation of Peking opera face to realize personalization.Based on Laplacian algorithm and BlendShape algorithm,this paper studies the personalized face of Peking Opera.This paper proposes a method to construct the shape base by local deformation of AU unit division of Peking opera face,which solves the problem of abnormal deformation of unrelated areas caused by Laplacian algorithm.Finally,a personalized interactive platform is built to map the face feature point data to a of Peking Opera Face AU unit to realize personalized Facebook.Experiments show that this method solves the problem of abnormal deformation,more in line with the performance effect of Peking Opera Face,real-time and efficient,natural expression.